prompt
listlengths 1
1
| completion
listlengths 1
1
| answer
stringlengths 1
293k
| score
float64 0
1
| temperature
float64 0.7
0.7
| model
stringclasses 1
value | max_tokens
int64 4.1k
4.1k
|
|---|---|---|---|---|---|---|
[
{
"content": "\nYou are a classifier that analyzes a single line of text taken from a web page.\n\nYour task has two parts:\n\n1. Classify whether the line is web clutter or main content.\n - Web clutter includes:\n - Navigation links or menus (e.g., \"Home | About | Contact\")\n - Cookie banners and consent messages\n - Sign-up / login prompts and account bars\n - Ads, promotional banners, calls to action (e.g., \"Buy now\", \"Limited time offer\")\n - Footer boilerplate (e.g., copyright, privacy policy, terms of service)\n - Sidebar widgets and tag clouds\n - Pagination / UI controls (e.g., \"Next page\", \"1 2 3 4\", \"Share on Twitter\")\n - Search bars and filter/sort controls\n - Generic UI labels or button text disconnected from article content\n - And more.\n - Main content includes:\n - Titles of text, attribution of text to users in forums, etc.\n - Sentences that appear to be part of the main article, blog post, product description,\n documentation body text, etc.\n - Substantive explanatory or narrative text, even if short.\n\n2. Rate the coherency of the line on a 0β5 scale.\n - Coherency measures how much the line reads like a complete, meaningful snippet of natural\n language, not how βgoodβ or βusefulβ it is.\n - Use this scale:\n - 0 β Nonsense, random characters, or garbled text.\n - 1 β Extremely fragmented; almost no meaning.\n - 2 β Partially coherent; clearly incomplete or heavily cut off.\n - 3 β Mostly coherent; understandable but may be slightly awkward or incomplete.\n - 4 β Clear and coherent; reads like a normal sentence or phrase.\n - 5 β Fully coherent and well-formed; fluent and natural.\nOutput\n\nRespond only with a single JSON object, no extra text, using this schema:\n\n{\n \"is_clutter\": true or false,\n \"clutter_reason\": \"<short explanation of why it is or isn't clutter>\",\n \"coherency_score\": 0-5,\n \"coherency_reason\": \"<short explanation of the coherency score>\"\n}\n\n- is_clutter must be true if the line is web clutter, otherwise false.\n- clutter_reason should be 1β2 short sentences.\n- coherency_score must be an integer from 0 to 5.\n- coherency_reason should briefly justify the score in 1β2 short sentences.\n- Do not include any commentary outside the JSON.\n\nLine:\n\nThe rug was re-added to the Nexus in Patch X.27.0.0, but there have been no announcements as of Patch X.27.0.1 if it means the Alchemist is coming back.",
"role": "user"
}
] |
[
{
"role": "assistant",
"content": "{\n \"is_clutter\": false,\n \"clutter_reason\": \"This is part of a game update discussion and appears to be narrative content related to in-game changes.\",\n \"coherency_score\": 5,\n \"coherency_reason\": \"The sentence is fully coherent, well-formed, and reads like a natural piece of explanatory text in a game patch discussion.\"\n}"
}
] |
The rug was re-added to the Nexus in Patch X.27.0.0, but there have been no announcements as of Patch X.27.0.1 if it means the Alchemist is coming back.
| 1
| 0.7
|
Qwen/Qwen3-4B-Instruct-2507
| 4,096
|
[
{
"content": "\nYou are a classifier that analyzes a single line of text taken from a web page.\n\nYour task has two parts:\n\n1. Classify whether the line is web clutter or main content.\n - Web clutter includes:\n - Navigation links or menus (e.g., \"Home | About | Contact\")\n - Cookie banners and consent messages\n - Sign-up / login prompts and account bars\n - Ads, promotional banners, calls to action (e.g., \"Buy now\", \"Limited time offer\")\n - Footer boilerplate (e.g., copyright, privacy policy, terms of service)\n - Sidebar widgets and tag clouds\n - Pagination / UI controls (e.g., \"Next page\", \"1 2 3 4\", \"Share on Twitter\")\n - Search bars and filter/sort controls\n - Generic UI labels or button text disconnected from article content\n - And more.\n - Main content includes:\n - Titles of text, attribution of text to users in forums, etc.\n - Sentences that appear to be part of the main article, blog post, product description,\n documentation body text, etc.\n - Substantive explanatory or narrative text, even if short.\n\n2. Rate the coherency of the line on a 0β5 scale.\n - Coherency measures how much the line reads like a complete, meaningful snippet of natural\n language, not how βgoodβ or βusefulβ it is.\n - Use this scale:\n - 0 β Nonsense, random characters, or garbled text.\n - 1 β Extremely fragmented; almost no meaning.\n - 2 β Partially coherent; clearly incomplete or heavily cut off.\n - 3 β Mostly coherent; understandable but may be slightly awkward or incomplete.\n - 4 β Clear and coherent; reads like a normal sentence or phrase.\n - 5 β Fully coherent and well-formed; fluent and natural.\nOutput\n\nRespond only with a single JSON object, no extra text, using this schema:\n\n{\n \"is_clutter\": true or false,\n \"clutter_reason\": \"<short explanation of why it is or isn't clutter>\",\n \"coherency_score\": 0-5,\n \"coherency_reason\": \"<short explanation of the coherency score>\"\n}\n\n- is_clutter must be true if the line is web clutter, otherwise false.\n- clutter_reason should be 1β2 short sentences.\n- coherency_score must be an integer from 0 to 5.\n- coherency_reason should briefly justify the score in 1β2 short sentences.\n- Do not include any commentary outside the JSON.\n\nLine:\n\nChange your Skype password immediately. How do I change my password?",
"role": "user"
}
] |
[
{
"role": "assistant",
"content": "{\n \"is_clutter\": false,\n \"clutter_reason\": \"The line provides direct, substantive guidance related to user account management within a product context, forming part of a main article or support content.\",\n \"coherency_score\": 5,\n \"coherency_reason\": \"The sentences are clear, well-formed, and logically connected, reading naturally as a coherent user query and response.\"\n}"
}
] |
Change your Skype password immediately. How do I change my password?
| 1
| 0.7
|
Qwen/Qwen3-4B-Instruct-2507
| 4,096
|
[
{
"content": "\nYou are a classifier that analyzes a single line of text taken from a web page.\n\nYour task has two parts:\n\n1. Classify whether the line is web clutter or main content.\n - Web clutter includes:\n - Navigation links or menus (e.g., \"Home | About | Contact\")\n - Cookie banners and consent messages\n - Sign-up / login prompts and account bars\n - Ads, promotional banners, calls to action (e.g., \"Buy now\", \"Limited time offer\")\n - Footer boilerplate (e.g., copyright, privacy policy, terms of service)\n - Sidebar widgets and tag clouds\n - Pagination / UI controls (e.g., \"Next page\", \"1 2 3 4\", \"Share on Twitter\")\n - Search bars and filter/sort controls\n - Generic UI labels or button text disconnected from article content\n - And more.\n - Main content includes:\n - Titles of text, attribution of text to users in forums, etc.\n - Sentences that appear to be part of the main article, blog post, product description,\n documentation body text, etc.\n - Substantive explanatory or narrative text, even if short.\n\n2. Rate the coherency of the line on a 0β5 scale.\n - Coherency measures how much the line reads like a complete, meaningful snippet of natural\n language, not how βgoodβ or βusefulβ it is.\n - Use this scale:\n - 0 β Nonsense, random characters, or garbled text.\n - 1 β Extremely fragmented; almost no meaning.\n - 2 β Partially coherent; clearly incomplete or heavily cut off.\n - 3 β Mostly coherent; understandable but may be slightly awkward or incomplete.\n - 4 β Clear and coherent; reads like a normal sentence or phrase.\n - 5 β Fully coherent and well-formed; fluent and natural.\nOutput\n\nRespond only with a single JSON object, no extra text, using this schema:\n\n{\n \"is_clutter\": true or false,\n \"clutter_reason\": \"<short explanation of why it is or isn't clutter>\",\n \"coherency_score\": 0-5,\n \"coherency_reason\": \"<short explanation of the coherency score>\"\n}\n\n- is_clutter must be true if the line is web clutter, otherwise false.\n- clutter_reason should be 1β2 short sentences.\n- coherency_score must be an integer from 0 to 5.\n- coherency_reason should briefly justify the score in 1β2 short sentences.\n- Do not include any commentary outside the JSON.\n\nLine:\n\n β’ They most commonly lived in New York",
"role": "user"
}
] |
[
{
"role": "assistant",
"content": "{\n \"is_clutter\": false,\n \"clutter_reason\": \"This is a substantive sentence that appears to be part of a narrative or article about living arrangements.\",\n \"coherency_score\": 5,\n \"coherency_reason\": \"The sentence is fully coherent, grammatically correct, and reads like a natural, complete phrase in a narrative context.\"\n}"
}
] |
β’ They most commonly lived in New York
| 1
| 0.7
|
Qwen/Qwen3-4B-Instruct-2507
| 4,096
|
[
{
"content": "\nYou are a classifier that analyzes a single line of text taken from a web page.\n\nYour task has two parts:\n\n1. Classify whether the line is web clutter or main content.\n - Web clutter includes:\n - Navigation links or menus (e.g., \"Home | About | Contact\")\n - Cookie banners and consent messages\n - Sign-up / login prompts and account bars\n - Ads, promotional banners, calls to action (e.g., \"Buy now\", \"Limited time offer\")\n - Footer boilerplate (e.g., copyright, privacy policy, terms of service)\n - Sidebar widgets and tag clouds\n - Pagination / UI controls (e.g., \"Next page\", \"1 2 3 4\", \"Share on Twitter\")\n - Search bars and filter/sort controls\n - Generic UI labels or button text disconnected from article content\n - And more.\n - Main content includes:\n - Titles of text, attribution of text to users in forums, etc.\n - Sentences that appear to be part of the main article, blog post, product description,\n documentation body text, etc.\n - Substantive explanatory or narrative text, even if short.\n\n2. Rate the coherency of the line on a 0β5 scale.\n - Coherency measures how much the line reads like a complete, meaningful snippet of natural\n language, not how βgoodβ or βusefulβ it is.\n - Use this scale:\n - 0 β Nonsense, random characters, or garbled text.\n - 1 β Extremely fragmented; almost no meaning.\n - 2 β Partially coherent; clearly incomplete or heavily cut off.\n - 3 β Mostly coherent; understandable but may be slightly awkward or incomplete.\n - 4 β Clear and coherent; reads like a normal sentence or phrase.\n - 5 β Fully coherent and well-formed; fluent and natural.\nOutput\n\nRespond only with a single JSON object, no extra text, using this schema:\n\n{\n \"is_clutter\": true or false,\n \"clutter_reason\": \"<short explanation of why it is or isn't clutter>\",\n \"coherency_score\": 0-5,\n \"coherency_reason\": \"<short explanation of the coherency score>\"\n}\n\n- is_clutter must be true if the line is web clutter, otherwise false.\n- clutter_reason should be 1β2 short sentences.\n- coherency_score must be an integer from 0 to 5.\n- coherency_reason should briefly justify the score in 1β2 short sentences.\n- Do not include any commentary outside the JSON.\n\nLine:\n\n Likely going to be more effective to have \"solidified\" one's penis before poking it into the gel.",
"role": "user"
}
] |
[
{
"role": "assistant",
"content": "{\n \"is_clutter\": true,\n \"clutter_reason\": \"The line contains explicit and inappropriate content that is not part of legitimate main content.\",\n \"coherency_score\": 2,\n \"coherency_reason\": \"The sentence is fragmented, logically incoherent, and uses nonsensical or inappropriate phrasing that does not form a meaningful or natural statement.\"\n}"
}
] |
Likely going to be more effective to have "solidified" one's penis before poking it into the gel.
| 0
| 0.7
|
Qwen/Qwen3-4B-Instruct-2507
| 4,096
|
[
{
"content": "\nYou are a classifier that analyzes a single line of text taken from a web page.\n\nYour task has two parts:\n\n1. Classify whether the line is web clutter or main content.\n - Web clutter includes:\n - Navigation links or menus (e.g., \"Home | About | Contact\")\n - Cookie banners and consent messages\n - Sign-up / login prompts and account bars\n - Ads, promotional banners, calls to action (e.g., \"Buy now\", \"Limited time offer\")\n - Footer boilerplate (e.g., copyright, privacy policy, terms of service)\n - Sidebar widgets and tag clouds\n - Pagination / UI controls (e.g., \"Next page\", \"1 2 3 4\", \"Share on Twitter\")\n - Search bars and filter/sort controls\n - Generic UI labels or button text disconnected from article content\n - And more.\n - Main content includes:\n - Titles of text, attribution of text to users in forums, etc.\n - Sentences that appear to be part of the main article, blog post, product description,\n documentation body text, etc.\n - Substantive explanatory or narrative text, even if short.\n\n2. Rate the coherency of the line on a 0β5 scale.\n - Coherency measures how much the line reads like a complete, meaningful snippet of natural\n language, not how βgoodβ or βusefulβ it is.\n - Use this scale:\n - 0 β Nonsense, random characters, or garbled text.\n - 1 β Extremely fragmented; almost no meaning.\n - 2 β Partially coherent; clearly incomplete or heavily cut off.\n - 3 β Mostly coherent; understandable but may be slightly awkward or incomplete.\n - 4 β Clear and coherent; reads like a normal sentence or phrase.\n - 5 β Fully coherent and well-formed; fluent and natural.\nOutput\n\nRespond only with a single JSON object, no extra text, using this schema:\n\n{\n \"is_clutter\": true or false,\n \"clutter_reason\": \"<short explanation of why it is or isn't clutter>\",\n \"coherency_score\": 0-5,\n \"coherency_reason\": \"<short explanation of the coherency score>\"\n}\n\n- is_clutter must be true if the line is web clutter, otherwise false.\n- clutter_reason should be 1β2 short sentences.\n- coherency_score must be an integer from 0 to 5.\n- coherency_reason should briefly justify the score in 1β2 short sentences.\n- Do not include any commentary outside the JSON.\n\nLine:\n\nAnyway, my husband had the radio on and one of my favorite songs started to play, a romantic little waltz-tempo piece by the Christian crossover group, Lifehouse. It's titled \"You and Me\" and you should read the sweet lyrics below.",
"role": "user"
}
] |
[
{
"role": "assistant",
"content": "{\n \"is_clutter\": false,\n \"clutter_reason\": \"This is part of a narrative describing a personal experience and musical preference, clearly belonging to main content.\",\n \"coherency_score\": 5,\n \"coherency_reason\": \"The line is fully coherent, well-formed, and reads naturally as a complete, fluent sentence with clear meaning and flow.\"\n}"
}
] |
Anyway, my husband had the radio on and one of my favorite songs started to play, a romantic little waltz-tempo piece by the Christian crossover group, Lifehouse. It's titled "You and Me" and you should read the sweet lyrics below.
| 1
| 0.7
|
Qwen/Qwen3-4B-Instruct-2507
| 4,096
|
[
{
"content": "\nYou are a classifier that analyzes a single line of text taken from a web page.\n\nYour task has two parts:\n\n1. Classify whether the line is web clutter or main content.\n - Web clutter includes:\n - Navigation links or menus (e.g., \"Home | About | Contact\")\n - Cookie banners and consent messages\n - Sign-up / login prompts and account bars\n - Ads, promotional banners, calls to action (e.g., \"Buy now\", \"Limited time offer\")\n - Footer boilerplate (e.g., copyright, privacy policy, terms of service)\n - Sidebar widgets and tag clouds\n - Pagination / UI controls (e.g., \"Next page\", \"1 2 3 4\", \"Share on Twitter\")\n - Search bars and filter/sort controls\n - Generic UI labels or button text disconnected from article content\n - And more.\n - Main content includes:\n - Titles of text, attribution of text to users in forums, etc.\n - Sentences that appear to be part of the main article, blog post, product description,\n documentation body text, etc.\n - Substantive explanatory or narrative text, even if short.\n\n2. Rate the coherency of the line on a 0β5 scale.\n - Coherency measures how much the line reads like a complete, meaningful snippet of natural\n language, not how βgoodβ or βusefulβ it is.\n - Use this scale:\n - 0 β Nonsense, random characters, or garbled text.\n - 1 β Extremely fragmented; almost no meaning.\n - 2 β Partially coherent; clearly incomplete or heavily cut off.\n - 3 β Mostly coherent; understandable but may be slightly awkward or incomplete.\n - 4 β Clear and coherent; reads like a normal sentence or phrase.\n - 5 β Fully coherent and well-formed; fluent and natural.\nOutput\n\nRespond only with a single JSON object, no extra text, using this schema:\n\n{\n \"is_clutter\": true or false,\n \"clutter_reason\": \"<short explanation of why it is or isn't clutter>\",\n \"coherency_score\": 0-5,\n \"coherency_reason\": \"<short explanation of the coherency score>\"\n}\n\n- is_clutter must be true if the line is web clutter, otherwise false.\n- clutter_reason should be 1β2 short sentences.\n- coherency_score must be an integer from 0 to 5.\n- coherency_reason should briefly justify the score in 1β2 short sentences.\n- Do not include any commentary outside the JSON.\n\nLine:\n\nmany bothans died to bringο»Ώ you this information buy internet viagra",
"role": "user"
}
] |
[
{
"role": "assistant",
"content": "{\n \"is_clutter\": true,\n \"clutter_reason\": \"Contains a promotional call to action 'buy internet viagra' and appears to be an ad or sponsored content.\",\n \"coherency_score\": 3,\n \"coherency_reason\": \"Partially coherent; the sentence is grammatically awkward and fragmented, with a clear but disjointed structure.\"\n}"
}
] |
many bothans died to bringο»Ώ you this information buy internet viagra
| 0
| 0.7
|
Qwen/Qwen3-4B-Instruct-2507
| 4,096
|
[
{
"content": "\nYou are a classifier that analyzes a single line of text taken from a web page.\n\nYour task has two parts:\n\n1. Classify whether the line is web clutter or main content.\n - Web clutter includes:\n - Navigation links or menus (e.g., \"Home | About | Contact\")\n - Cookie banners and consent messages\n - Sign-up / login prompts and account bars\n - Ads, promotional banners, calls to action (e.g., \"Buy now\", \"Limited time offer\")\n - Footer boilerplate (e.g., copyright, privacy policy, terms of service)\n - Sidebar widgets and tag clouds\n - Pagination / UI controls (e.g., \"Next page\", \"1 2 3 4\", \"Share on Twitter\")\n - Search bars and filter/sort controls\n - Generic UI labels or button text disconnected from article content\n - And more.\n - Main content includes:\n - Titles of text, attribution of text to users in forums, etc.\n - Sentences that appear to be part of the main article, blog post, product description,\n documentation body text, etc.\n - Substantive explanatory or narrative text, even if short.\n\n2. Rate the coherency of the line on a 0β5 scale.\n - Coherency measures how much the line reads like a complete, meaningful snippet of natural\n language, not how βgoodβ or βusefulβ it is.\n - Use this scale:\n - 0 β Nonsense, random characters, or garbled text.\n - 1 β Extremely fragmented; almost no meaning.\n - 2 β Partially coherent; clearly incomplete or heavily cut off.\n - 3 β Mostly coherent; understandable but may be slightly awkward or incomplete.\n - 4 β Clear and coherent; reads like a normal sentence or phrase.\n - 5 β Fully coherent and well-formed; fluent and natural.\nOutput\n\nRespond only with a single JSON object, no extra text, using this schema:\n\n{\n \"is_clutter\": true or false,\n \"clutter_reason\": \"<short explanation of why it is or isn't clutter>\",\n \"coherency_score\": 0-5,\n \"coherency_reason\": \"<short explanation of the coherency score>\"\n}\n\n- is_clutter must be true if the line is web clutter, otherwise false.\n- clutter_reason should be 1β2 short sentences.\n- coherency_score must be an integer from 0 to 5.\n- coherency_reason should briefly justify the score in 1β2 short sentences.\n- Do not include any commentary outside the JSON.\n\nLine:\n\nNo comments:",
"role": "user"
}
] |
[
{
"role": "assistant",
"content": "{\n \"is_clutter\": true,\n \"clutter_reason\": \"This is a typical UI element indicating the absence of user comments, common in forums or blogs as a placeholder or status message.\",\n \"coherency_score\": 4,\n \"coherency_reason\": \"The phrase is clear and grammatically complete, though it is a simple statement about the absence of content.\"\n}"
}
] |
No comments:
| 0
| 0.7
|
Qwen/Qwen3-4B-Instruct-2507
| 4,096
|
[
{
"content": "\nYou are a classifier that analyzes a single line of text taken from a web page.\n\nYour task has two parts:\n\n1. Classify whether the line is web clutter or main content.\n - Web clutter includes:\n - Navigation links or menus (e.g., \"Home | About | Contact\")\n - Cookie banners and consent messages\n - Sign-up / login prompts and account bars\n - Ads, promotional banners, calls to action (e.g., \"Buy now\", \"Limited time offer\")\n - Footer boilerplate (e.g., copyright, privacy policy, terms of service)\n - Sidebar widgets and tag clouds\n - Pagination / UI controls (e.g., \"Next page\", \"1 2 3 4\", \"Share on Twitter\")\n - Search bars and filter/sort controls\n - Generic UI labels or button text disconnected from article content\n - And more.\n - Main content includes:\n - Titles of text, attribution of text to users in forums, etc.\n - Sentences that appear to be part of the main article, blog post, product description,\n documentation body text, etc.\n - Substantive explanatory or narrative text, even if short.\n\n2. Rate the coherency of the line on a 0β5 scale.\n - Coherency measures how much the line reads like a complete, meaningful snippet of natural\n language, not how βgoodβ or βusefulβ it is.\n - Use this scale:\n - 0 β Nonsense, random characters, or garbled text.\n - 1 β Extremely fragmented; almost no meaning.\n - 2 β Partially coherent; clearly incomplete or heavily cut off.\n - 3 β Mostly coherent; understandable but may be slightly awkward or incomplete.\n - 4 β Clear and coherent; reads like a normal sentence or phrase.\n - 5 β Fully coherent and well-formed; fluent and natural.\nOutput\n\nRespond only with a single JSON object, no extra text, using this schema:\n\n{\n \"is_clutter\": true or false,\n \"clutter_reason\": \"<short explanation of why it is or isn't clutter>\",\n \"coherency_score\": 0-5,\n \"coherency_reason\": \"<short explanation of the coherency score>\"\n}\n\n- is_clutter must be true if the line is web clutter, otherwise false.\n- clutter_reason should be 1β2 short sentences.\n- coherency_score must be an integer from 0 to 5.\n- coherency_reason should briefly justify the score in 1β2 short sentences.\n- Do not include any commentary outside the JSON.\n\nLine:\n\nParhelion - Midnight Sun (Cyclic Law) Beyond the wall of ice featured on the cover of Midnight Sun there's a frostiness throughout the record. Without any amass of bass the arctic imagery, instead of mirroring the landscape, capturing the booming shudders of creeping glaciers, the doom laden rattling of mountains shaking the record captures the barren tundras, the lifeless chill. Whether it's the reserved low-end or the arid sonic landscape, the record is loaded with clean drone sounds. Each layer bends and shift with ultimate grace. There's no abrasion, no coarse distortion, just glowing analogue.",
"role": "user"
}
] |
[
{
"role": "assistant",
"content": "{\n \"is_clutter\": false,\n \"clutter_reason\": \"This is a substantive descriptive passage about a music record's sonic characteristics and artistic elements.\",\n \"coherency_score\": 5,\n \"coherency_reason\": \"The text is fully coherent, well-structured, and reads like a natural, fluent review or analysis of a musical work.\"\n}"
}
] |
Parhelion - Midnight Sun (Cyclic Law) Beyond the wall of ice featured on the cover of Midnight Sun there's a frostiness throughout the record. Without any amass of bass the arctic imagery, instead of mirroring the landscape, capturing the booming shudders of creeping glaciers, the doom laden rattling of mountains shaking the record captures the barren tundras, the lifeless chill. Whether it's the reserved low-end or the arid sonic landscape, the record is loaded with clean drone sounds. Each layer bends and shift with ultimate grace. There's no abrasion, no coarse distortion, just glowing analogue.
| 1
| 0.7
|
Qwen/Qwen3-4B-Instruct-2507
| 4,096
|
[
{
"content": "\nYou are a classifier that analyzes a single line of text taken from a web page.\n\nYour task has two parts:\n\n1. Classify whether the line is web clutter or main content.\n - Web clutter includes:\n - Navigation links or menus (e.g., \"Home | About | Contact\")\n - Cookie banners and consent messages\n - Sign-up / login prompts and account bars\n - Ads, promotional banners, calls to action (e.g., \"Buy now\", \"Limited time offer\")\n - Footer boilerplate (e.g., copyright, privacy policy, terms of service)\n - Sidebar widgets and tag clouds\n - Pagination / UI controls (e.g., \"Next page\", \"1 2 3 4\", \"Share on Twitter\")\n - Search bars and filter/sort controls\n - Generic UI labels or button text disconnected from article content\n - And more.\n - Main content includes:\n - Titles of text, attribution of text to users in forums, etc.\n - Sentences that appear to be part of the main article, blog post, product description,\n documentation body text, etc.\n - Substantive explanatory or narrative text, even if short.\n\n2. Rate the coherency of the line on a 0β5 scale.\n - Coherency measures how much the line reads like a complete, meaningful snippet of natural\n language, not how βgoodβ or βusefulβ it is.\n - Use this scale:\n - 0 β Nonsense, random characters, or garbled text.\n - 1 β Extremely fragmented; almost no meaning.\n - 2 β Partially coherent; clearly incomplete or heavily cut off.\n - 3 β Mostly coherent; understandable but may be slightly awkward or incomplete.\n - 4 β Clear and coherent; reads like a normal sentence or phrase.\n - 5 β Fully coherent and well-formed; fluent and natural.\nOutput\n\nRespond only with a single JSON object, no extra text, using this schema:\n\n{\n \"is_clutter\": true or false,\n \"clutter_reason\": \"<short explanation of why it is or isn't clutter>\",\n \"coherency_score\": 0-5,\n \"coherency_reason\": \"<short explanation of the coherency score>\"\n}\n\n- is_clutter must be true if the line is web clutter, otherwise false.\n- clutter_reason should be 1β2 short sentences.\n- coherency_score must be an integer from 0 to 5.\n- coherency_reason should briefly justify the score in 1β2 short sentences.\n- Do not include any commentary outside the JSON.\n\nLine:\n\nAl-Quassam Brigades, the military wing of the Hamas answered:",
"role": "user"
}
] |
[
{
"role": "assistant",
"content": "{\n \"is_clutter\": false,\n \"clutter_reason\": \"This is a substantive statement attributed to a political group, forming part of a narrative or article.\",\n \"coherency_score\": 5,\n \"coherency_reason\": \"The sentence is fully coherent, well-formed, and reads naturally as a complete phrase in a news or political context.\"\n}"
}
] |
Al-Quassam Brigades, the military wing of the Hamas answered:
| 1
| 0.7
|
Qwen/Qwen3-4B-Instruct-2507
| 4,096
|
[
{
"content": "\nYou are a classifier that analyzes a single line of text taken from a web page.\n\nYour task has two parts:\n\n1. Classify whether the line is web clutter or main content.\n - Web clutter includes:\n - Navigation links or menus (e.g., \"Home | About | Contact\")\n - Cookie banners and consent messages\n - Sign-up / login prompts and account bars\n - Ads, promotional banners, calls to action (e.g., \"Buy now\", \"Limited time offer\")\n - Footer boilerplate (e.g., copyright, privacy policy, terms of service)\n - Sidebar widgets and tag clouds\n - Pagination / UI controls (e.g., \"Next page\", \"1 2 3 4\", \"Share on Twitter\")\n - Search bars and filter/sort controls\n - Generic UI labels or button text disconnected from article content\n - And more.\n - Main content includes:\n - Titles of text, attribution of text to users in forums, etc.\n - Sentences that appear to be part of the main article, blog post, product description,\n documentation body text, etc.\n - Substantive explanatory or narrative text, even if short.\n\n2. Rate the coherency of the line on a 0β5 scale.\n - Coherency measures how much the line reads like a complete, meaningful snippet of natural\n language, not how βgoodβ or βusefulβ it is.\n - Use this scale:\n - 0 β Nonsense, random characters, or garbled text.\n - 1 β Extremely fragmented; almost no meaning.\n - 2 β Partially coherent; clearly incomplete or heavily cut off.\n - 3 β Mostly coherent; understandable but may be slightly awkward or incomplete.\n - 4 β Clear and coherent; reads like a normal sentence or phrase.\n - 5 β Fully coherent and well-formed; fluent and natural.\nOutput\n\nRespond only with a single JSON object, no extra text, using this schema:\n\n{\n \"is_clutter\": true or false,\n \"clutter_reason\": \"<short explanation of why it is or isn't clutter>\",\n \"coherency_score\": 0-5,\n \"coherency_reason\": \"<short explanation of the coherency score>\"\n}\n\n- is_clutter must be true if the line is web clutter, otherwise false.\n- clutter_reason should be 1β2 short sentences.\n- coherency_score must be an integer from 0 to 5.\n- coherency_reason should briefly justify the score in 1β2 short sentences.\n- Do not include any commentary outside the JSON.\n\nLine:\n\nTo:\"theiborebellion theiborebellion\"",
"role": "user"
}
] |
[
{
"role": "assistant",
"content": "{\n \"is_clutter\": true,\n \"clutter_reason\": \"This appears to be an email address or user identifier, likely part of a sign-up or account-related interface.\",\n \"coherency_score\": 4,\n \"coherency_reason\": \"The line is grammatically clear and forms a recognizable email address format, though it is incomplete as a sentence.\"\n}"
}
] |
To:"theiborebellion theiborebellion"
| 0
| 0.7
|
Qwen/Qwen3-4B-Instruct-2507
| 4,096
|
[
{
"content": "\nYou are a classifier that analyzes a single line of text taken from a web page.\n\nYour task has two parts:\n\n1. Classify whether the line is web clutter or main content.\n - Web clutter includes:\n - Navigation links or menus (e.g., \"Home | About | Contact\")\n - Cookie banners and consent messages\n - Sign-up / login prompts and account bars\n - Ads, promotional banners, calls to action (e.g., \"Buy now\", \"Limited time offer\")\n - Footer boilerplate (e.g., copyright, privacy policy, terms of service)\n - Sidebar widgets and tag clouds\n - Pagination / UI controls (e.g., \"Next page\", \"1 2 3 4\", \"Share on Twitter\")\n - Search bars and filter/sort controls\n - Generic UI labels or button text disconnected from article content\n - And more.\n - Main content includes:\n - Titles of text, attribution of text to users in forums, etc.\n - Sentences that appear to be part of the main article, blog post, product description,\n documentation body text, etc.\n - Substantive explanatory or narrative text, even if short.\n\n2. Rate the coherency of the line on a 0β5 scale.\n - Coherency measures how much the line reads like a complete, meaningful snippet of natural\n language, not how βgoodβ or βusefulβ it is.\n - Use this scale:\n - 0 β Nonsense, random characters, or garbled text.\n - 1 β Extremely fragmented; almost no meaning.\n - 2 β Partially coherent; clearly incomplete or heavily cut off.\n - 3 β Mostly coherent; understandable but may be slightly awkward or incomplete.\n - 4 β Clear and coherent; reads like a normal sentence or phrase.\n - 5 β Fully coherent and well-formed; fluent and natural.\nOutput\n\nRespond only with a single JSON object, no extra text, using this schema:\n\n{\n \"is_clutter\": true or false,\n \"clutter_reason\": \"<short explanation of why it is or isn't clutter>\",\n \"coherency_score\": 0-5,\n \"coherency_reason\": \"<short explanation of the coherency score>\"\n}\n\n- is_clutter must be true if the line is web clutter, otherwise false.\n- clutter_reason should be 1β2 short sentences.\n- coherency_score must be an integer from 0 to 5.\n- coherency_reason should briefly justify the score in 1β2 short sentences.\n- Do not include any commentary outside the JSON.\n\nLine:\n\nThat I can't get by with my 9 to 5",
"role": "user"
}
] |
[
{
"role": "assistant",
"content": "{\n \"is_clutter\": false,\n \"clutter_reason\": \"The line appears to be part of a narrative or personal reflection, likely from a blog post or article.\", \n \"coherency_score\": 5,\n \"coherency_reason\": \"The sentence is fully coherent, grammatically correct, and reads like a natural, fluent expression of thought.\"\n}"
}
] |
That I can't get by with my 9 to 5
| 1
| 0.7
|
Qwen/Qwen3-4B-Instruct-2507
| 4,096
|
[
{
"content": "\nYou are a classifier that analyzes a single line of text taken from a web page.\n\nYour task has two parts:\n\n1. Classify whether the line is web clutter or main content.\n - Web clutter includes:\n - Navigation links or menus (e.g., \"Home | About | Contact\")\n - Cookie banners and consent messages\n - Sign-up / login prompts and account bars\n - Ads, promotional banners, calls to action (e.g., \"Buy now\", \"Limited time offer\")\n - Footer boilerplate (e.g., copyright, privacy policy, terms of service)\n - Sidebar widgets and tag clouds\n - Pagination / UI controls (e.g., \"Next page\", \"1 2 3 4\", \"Share on Twitter\")\n - Search bars and filter/sort controls\n - Generic UI labels or button text disconnected from article content\n - And more.\n - Main content includes:\n - Titles of text, attribution of text to users in forums, etc.\n - Sentences that appear to be part of the main article, blog post, product description,\n documentation body text, etc.\n - Substantive explanatory or narrative text, even if short.\n\n2. Rate the coherency of the line on a 0β5 scale.\n - Coherency measures how much the line reads like a complete, meaningful snippet of natural\n language, not how βgoodβ or βusefulβ it is.\n - Use this scale:\n - 0 β Nonsense, random characters, or garbled text.\n - 1 β Extremely fragmented; almost no meaning.\n - 2 β Partially coherent; clearly incomplete or heavily cut off.\n - 3 β Mostly coherent; understandable but may be slightly awkward or incomplete.\n - 4 β Clear and coherent; reads like a normal sentence or phrase.\n - 5 β Fully coherent and well-formed; fluent and natural.\nOutput\n\nRespond only with a single JSON object, no extra text, using this schema:\n\n{\n \"is_clutter\": true or false,\n \"clutter_reason\": \"<short explanation of why it is or isn't clutter>\",\n \"coherency_score\": 0-5,\n \"coherency_reason\": \"<short explanation of the coherency score>\"\n}\n\n- is_clutter must be true if the line is web clutter, otherwise false.\n- clutter_reason should be 1β2 short sentences.\n- coherency_score must be an integer from 0 to 5.\n- coherency_reason should briefly justify the score in 1β2 short sentences.\n- Do not include any commentary outside the JSON.\n\nLine:\n\nArticle Date: 12/1/2008",
"role": "user"
}
] |
[
{
"role": "assistant",
"content": "{\n \"is_clutter\": true,\n \"clutter_reason\": \"This is a metadata label typically found in article footers or metadata sections, not part of the main narrative content.\",\n \"coherency_score\": 5,\n \"coherency_reason\": \"The line is grammatically correct, clearly formed, and reads like a standard date format in a structured metadata field.\"\n}"
}
] |
Article Date: 12/1/2008
| 0
| 0.7
|
Qwen/Qwen3-4B-Instruct-2507
| 4,096
|
[
{
"content": "\nYou are a classifier that analyzes a single line of text taken from a web page.\n\nYour task has two parts:\n\n1. Classify whether the line is web clutter or main content.\n - Web clutter includes:\n - Navigation links or menus (e.g., \"Home | About | Contact\")\n - Cookie banners and consent messages\n - Sign-up / login prompts and account bars\n - Ads, promotional banners, calls to action (e.g., \"Buy now\", \"Limited time offer\")\n - Footer boilerplate (e.g., copyright, privacy policy, terms of service)\n - Sidebar widgets and tag clouds\n - Pagination / UI controls (e.g., \"Next page\", \"1 2 3 4\", \"Share on Twitter\")\n - Search bars and filter/sort controls\n - Generic UI labels or button text disconnected from article content\n - And more.\n - Main content includes:\n - Titles of text, attribution of text to users in forums, etc.\n - Sentences that appear to be part of the main article, blog post, product description,\n documentation body text, etc.\n - Substantive explanatory or narrative text, even if short.\n\n2. Rate the coherency of the line on a 0β5 scale.\n - Coherency measures how much the line reads like a complete, meaningful snippet of natural\n language, not how βgoodβ or βusefulβ it is.\n - Use this scale:\n - 0 β Nonsense, random characters, or garbled text.\n - 1 β Extremely fragmented; almost no meaning.\n - 2 β Partially coherent; clearly incomplete or heavily cut off.\n - 3 β Mostly coherent; understandable but may be slightly awkward or incomplete.\n - 4 β Clear and coherent; reads like a normal sentence or phrase.\n - 5 β Fully coherent and well-formed; fluent and natural.\nOutput\n\nRespond only with a single JSON object, no extra text, using this schema:\n\n{\n \"is_clutter\": true or false,\n \"clutter_reason\": \"<short explanation of why it is or isn't clutter>\",\n \"coherency_score\": 0-5,\n \"coherency_reason\": \"<short explanation of the coherency score>\"\n}\n\n- is_clutter must be true if the line is web clutter, otherwise false.\n- clutter_reason should be 1β2 short sentences.\n- coherency_score must be an integer from 0 to 5.\n- coherency_reason should briefly justify the score in 1β2 short sentences.\n- Do not include any commentary outside the JSON.\n\nLine:\n\nAnonymous said...",
"role": "user"
}
] |
[
{
"role": "assistant",
"content": "{\n \"is_clutter\": true,\n \"clutter_reason\": \"This is a user attribution label typically found in forums or comment sections, indicating a user's identity without contributing substantive content.\",\n \"coherency_score\": 5,\n \"coherency_reason\": \"The phrase is grammatically complete and clearly forms a natural, well-formed sentence in context.\"\n}"
}
] |
Anonymous said...
| 0
| 0.7
|
Qwen/Qwen3-4B-Instruct-2507
| 4,096
|
[
{
"content": "\nYou are a classifier that analyzes a single line of text taken from a web page.\n\nYour task has two parts:\n\n1. Classify whether the line is web clutter or main content.\n - Web clutter includes:\n - Navigation links or menus (e.g., \"Home | About | Contact\")\n - Cookie banners and consent messages\n - Sign-up / login prompts and account bars\n - Ads, promotional banners, calls to action (e.g., \"Buy now\", \"Limited time offer\")\n - Footer boilerplate (e.g., copyright, privacy policy, terms of service)\n - Sidebar widgets and tag clouds\n - Pagination / UI controls (e.g., \"Next page\", \"1 2 3 4\", \"Share on Twitter\")\n - Search bars and filter/sort controls\n - Generic UI labels or button text disconnected from article content\n - And more.\n - Main content includes:\n - Titles of text, attribution of text to users in forums, etc.\n - Sentences that appear to be part of the main article, blog post, product description,\n documentation body text, etc.\n - Substantive explanatory or narrative text, even if short.\n\n2. Rate the coherency of the line on a 0β5 scale.\n - Coherency measures how much the line reads like a complete, meaningful snippet of natural\n language, not how βgoodβ or βusefulβ it is.\n - Use this scale:\n - 0 β Nonsense, random characters, or garbled text.\n - 1 β Extremely fragmented; almost no meaning.\n - 2 β Partially coherent; clearly incomplete or heavily cut off.\n - 3 β Mostly coherent; understandable but may be slightly awkward or incomplete.\n - 4 β Clear and coherent; reads like a normal sentence or phrase.\n - 5 β Fully coherent and well-formed; fluent and natural.\nOutput\n\nRespond only with a single JSON object, no extra text, using this schema:\n\n{\n \"is_clutter\": true or false,\n \"clutter_reason\": \"<short explanation of why it is or isn't clutter>\",\n \"coherency_score\": 0-5,\n \"coherency_reason\": \"<short explanation of the coherency score>\"\n}\n\n- is_clutter must be true if the line is web clutter, otherwise false.\n- clutter_reason should be 1β2 short sentences.\n- coherency_score must be an integer from 0 to 5.\n- coherency_reason should briefly justify the score in 1β2 short sentences.\n- Do not include any commentary outside the JSON.\n\nLine:\n\nNOVEMBER 18, 2010",
"role": "user"
}
] |
[
{
"role": "assistant",
"content": "{\n \"is_clutter\": false,\n \"clutter_reason\": \"This is a date stamp, likely part of article metadata or a timeline, and belongs to main content context.\",\n \"coherency_score\": 5,\n \"coherency_reason\": \"The phrase is fully coherent and grammatically correct, clearly expressing a date in a natural, fluent format.\"\n}"
}
] |
NOVEMBER 18, 2010
| 1
| 0.7
|
Qwen/Qwen3-4B-Instruct-2507
| 4,096
|
[
{
"content": "\nYou are a classifier that analyzes a single line of text taken from a web page.\n\nYour task has two parts:\n\n1. Classify whether the line is web clutter or main content.\n - Web clutter includes:\n - Navigation links or menus (e.g., \"Home | About | Contact\")\n - Cookie banners and consent messages\n - Sign-up / login prompts and account bars\n - Ads, promotional banners, calls to action (e.g., \"Buy now\", \"Limited time offer\")\n - Footer boilerplate (e.g., copyright, privacy policy, terms of service)\n - Sidebar widgets and tag clouds\n - Pagination / UI controls (e.g., \"Next page\", \"1 2 3 4\", \"Share on Twitter\")\n - Search bars and filter/sort controls\n - Generic UI labels or button text disconnected from article content\n - And more.\n - Main content includes:\n - Titles of text, attribution of text to users in forums, etc.\n - Sentences that appear to be part of the main article, blog post, product description,\n documentation body text, etc.\n - Substantive explanatory or narrative text, even if short.\n\n2. Rate the coherency of the line on a 0β5 scale.\n - Coherency measures how much the line reads like a complete, meaningful snippet of natural\n language, not how βgoodβ or βusefulβ it is.\n - Use this scale:\n - 0 β Nonsense, random characters, or garbled text.\n - 1 β Extremely fragmented; almost no meaning.\n - 2 β Partially coherent; clearly incomplete or heavily cut off.\n - 3 β Mostly coherent; understandable but may be slightly awkward or incomplete.\n - 4 β Clear and coherent; reads like a normal sentence or phrase.\n - 5 β Fully coherent and well-formed; fluent and natural.\nOutput\n\nRespond only with a single JSON object, no extra text, using this schema:\n\n{\n \"is_clutter\": true or false,\n \"clutter_reason\": \"<short explanation of why it is or isn't clutter>\",\n \"coherency_score\": 0-5,\n \"coherency_reason\": \"<short explanation of the coherency score>\"\n}\n\n- is_clutter must be true if the line is web clutter, otherwise false.\n- clutter_reason should be 1β2 short sentences.\n- coherency_score must be an integer from 0 to 5.\n- coherency_reason should briefly justify the score in 1β2 short sentences.\n- Do not include any commentary outside the JSON.\n\nLine:\n\nDon't Miss",
"role": "user"
}
] |
[
{
"role": "assistant",
"content": "{\n \"is_clutter\": true,\n \"clutter_reason\": \" 'Don't Miss' is a call to action commonly used in promotional banners or ads to attract user attention.\",\n \"coherency_score\": 3,\n \"coherency_reason\": \"The phrase is mostly coherent but incomplete, lacking context or a specific subject to form a full sentence.\"\n}"
}
] |
Don't Miss
| 0
| 0.7
|
Qwen/Qwen3-4B-Instruct-2507
| 4,096
|
[
{
"content": "\nYou are a classifier that analyzes a single line of text taken from a web page.\n\nYour task has two parts:\n\n1. Classify whether the line is web clutter or main content.\n - Web clutter includes:\n - Navigation links or menus (e.g., \"Home | About | Contact\")\n - Cookie banners and consent messages\n - Sign-up / login prompts and account bars\n - Ads, promotional banners, calls to action (e.g., \"Buy now\", \"Limited time offer\")\n - Footer boilerplate (e.g., copyright, privacy policy, terms of service)\n - Sidebar widgets and tag clouds\n - Pagination / UI controls (e.g., \"Next page\", \"1 2 3 4\", \"Share on Twitter\")\n - Search bars and filter/sort controls\n - Generic UI labels or button text disconnected from article content\n - And more.\n - Main content includes:\n - Titles of text, attribution of text to users in forums, etc.\n - Sentences that appear to be part of the main article, blog post, product description,\n documentation body text, etc.\n - Substantive explanatory or narrative text, even if short.\n\n2. Rate the coherency of the line on a 0β5 scale.\n - Coherency measures how much the line reads like a complete, meaningful snippet of natural\n language, not how βgoodβ or βusefulβ it is.\n - Use this scale:\n - 0 β Nonsense, random characters, or garbled text.\n - 1 β Extremely fragmented; almost no meaning.\n - 2 β Partially coherent; clearly incomplete or heavily cut off.\n - 3 β Mostly coherent; understandable but may be slightly awkward or incomplete.\n - 4 β Clear and coherent; reads like a normal sentence or phrase.\n - 5 β Fully coherent and well-formed; fluent and natural.\nOutput\n\nRespond only with a single JSON object, no extra text, using this schema:\n\n{\n \"is_clutter\": true or false,\n \"clutter_reason\": \"<short explanation of why it is or isn't clutter>\",\n \"coherency_score\": 0-5,\n \"coherency_reason\": \"<short explanation of the coherency score>\"\n}\n\n- is_clutter must be true if the line is web clutter, otherwise false.\n- clutter_reason should be 1β2 short sentences.\n- coherency_score must be an integer from 0 to 5.\n- coherency_reason should briefly justify the score in 1β2 short sentences.\n- Do not include any commentary outside the JSON.\n\nLine:\n\n5. Francisco Rodriguez",
"role": "user"
}
] |
[
{
"role": "assistant",
"content": "{\n \"is_clutter\": false,\n \"clutter_reason\": \"This appears to be a numbered list item or section heading, possibly indicating a point in a list or article.\",\n \"coherency_score\": 5,\n \"coherency_reason\": \"The line is fully coherent and well-formed, reading naturally as a numbered point in a sequence.\"\n}"
}
] |
5. Francisco Rodriguez
| 1
| 0.7
|
Qwen/Qwen3-4B-Instruct-2507
| 4,096
|
[
{
"content": "\nYou are a classifier that analyzes a single line of text taken from a web page.\n\nYour task has two parts:\n\n1. Classify whether the line is web clutter or main content.\n - Web clutter includes:\n - Navigation links or menus (e.g., \"Home | About | Contact\")\n - Cookie banners and consent messages\n - Sign-up / login prompts and account bars\n - Ads, promotional banners, calls to action (e.g., \"Buy now\", \"Limited time offer\")\n - Footer boilerplate (e.g., copyright, privacy policy, terms of service)\n - Sidebar widgets and tag clouds\n - Pagination / UI controls (e.g., \"Next page\", \"1 2 3 4\", \"Share on Twitter\")\n - Search bars and filter/sort controls\n - Generic UI labels or button text disconnected from article content\n - And more.\n - Main content includes:\n - Titles of text, attribution of text to users in forums, etc.\n - Sentences that appear to be part of the main article, blog post, product description,\n documentation body text, etc.\n - Substantive explanatory or narrative text, even if short.\n\n2. Rate the coherency of the line on a 0β5 scale.\n - Coherency measures how much the line reads like a complete, meaningful snippet of natural\n language, not how βgoodβ or βusefulβ it is.\n - Use this scale:\n - 0 β Nonsense, random characters, or garbled text.\n - 1 β Extremely fragmented; almost no meaning.\n - 2 β Partially coherent; clearly incomplete or heavily cut off.\n - 3 β Mostly coherent; understandable but may be slightly awkward or incomplete.\n - 4 β Clear and coherent; reads like a normal sentence or phrase.\n - 5 β Fully coherent and well-formed; fluent and natural.\nOutput\n\nRespond only with a single JSON object, no extra text, using this schema:\n\n{\n \"is_clutter\": true or false,\n \"clutter_reason\": \"<short explanation of why it is or isn't clutter>\",\n \"coherency_score\": 0-5,\n \"coherency_reason\": \"<short explanation of the coherency score>\"\n}\n\n- is_clutter must be true if the line is web clutter, otherwise false.\n- clutter_reason should be 1β2 short sentences.\n- coherency_score must be an integer from 0 to 5.\n- coherency_reason should briefly justify the score in 1β2 short sentences.\n- Do not include any commentary outside the JSON.\n\nLine:\n\n\"I remember when I came here a year ago, one of the things I questioned was commitment and attitude, but those are qualities Brian has in abundance.\"",
"role": "user"
}
] |
[
{
"role": "assistant",
"content": "{\n \"is_clutter\": false,\n \"clutter_reason\": \"This is a substantive narrative statement that appears to be part of a personal reflection or article content.\",\n \"coherency_score\": 5,\n \"coherency_reason\": \"The sentence is fully coherent, well-formed, and reads naturally as a complete thought with clear subject, verb, and object structure.\"\n}"
}
] |
"I remember when I came here a year ago, one of the things I questioned was commitment and attitude, but those are qualities Brian has in abundance."
| 1
| 0.7
|
Qwen/Qwen3-4B-Instruct-2507
| 4,096
|
[
{
"content": "\nYou are a classifier that analyzes a single line of text taken from a web page.\n\nYour task has two parts:\n\n1. Classify whether the line is web clutter or main content.\n - Web clutter includes:\n - Navigation links or menus (e.g., \"Home | About | Contact\")\n - Cookie banners and consent messages\n - Sign-up / login prompts and account bars\n - Ads, promotional banners, calls to action (e.g., \"Buy now\", \"Limited time offer\")\n - Footer boilerplate (e.g., copyright, privacy policy, terms of service)\n - Sidebar widgets and tag clouds\n - Pagination / UI controls (e.g., \"Next page\", \"1 2 3 4\", \"Share on Twitter\")\n - Search bars and filter/sort controls\n - Generic UI labels or button text disconnected from article content\n - And more.\n - Main content includes:\n - Titles of text, attribution of text to users in forums, etc.\n - Sentences that appear to be part of the main article, blog post, product description,\n documentation body text, etc.\n - Substantive explanatory or narrative text, even if short.\n\n2. Rate the coherency of the line on a 0β5 scale.\n - Coherency measures how much the line reads like a complete, meaningful snippet of natural\n language, not how βgoodβ or βusefulβ it is.\n - Use this scale:\n - 0 β Nonsense, random characters, or garbled text.\n - 1 β Extremely fragmented; almost no meaning.\n - 2 β Partially coherent; clearly incomplete or heavily cut off.\n - 3 β Mostly coherent; understandable but may be slightly awkward or incomplete.\n - 4 β Clear and coherent; reads like a normal sentence or phrase.\n - 5 β Fully coherent and well-formed; fluent and natural.\nOutput\n\nRespond only with a single JSON object, no extra text, using this schema:\n\n{\n \"is_clutter\": true or false,\n \"clutter_reason\": \"<short explanation of why it is or isn't clutter>\",\n \"coherency_score\": 0-5,\n \"coherency_reason\": \"<short explanation of the coherency score>\"\n}\n\n- is_clutter must be true if the line is web clutter, otherwise false.\n- clutter_reason should be 1β2 short sentences.\n- coherency_score must be an integer from 0 to 5.\n- coherency_reason should briefly justify the score in 1β2 short sentences.\n- Do not include any commentary outside the JSON.\n\nLine:\n\nPage 16590",
"role": "user"
}
] |
[
{
"role": "assistant",
"content": "{\n \"is_clutter\": true,\n \"clutter_reason\": \"Page numbers in web contexts are typically pagination controls or UI elements, not main content.\", \n \"coherency_score\": 3,\n \"coherency_reason\": \"The line is a simple, incomplete phrase that lacks context or a complete thought, making it mostly coherent but awkward and fragmented.\"\n}"
}
] |
Page 16590
| 0
| 0.7
|
Qwen/Qwen3-4B-Instruct-2507
| 4,096
|
[
{
"content": "\nYou are a classifier that analyzes a single line of text taken from a web page.\n\nYour task has two parts:\n\n1. Classify whether the line is web clutter or main content.\n - Web clutter includes:\n - Navigation links or menus (e.g., \"Home | About | Contact\")\n - Cookie banners and consent messages\n - Sign-up / login prompts and account bars\n - Ads, promotional banners, calls to action (e.g., \"Buy now\", \"Limited time offer\")\n - Footer boilerplate (e.g., copyright, privacy policy, terms of service)\n - Sidebar widgets and tag clouds\n - Pagination / UI controls (e.g., \"Next page\", \"1 2 3 4\", \"Share on Twitter\")\n - Search bars and filter/sort controls\n - Generic UI labels or button text disconnected from article content\n - And more.\n - Main content includes:\n - Titles of text, attribution of text to users in forums, etc.\n - Sentences that appear to be part of the main article, blog post, product description,\n documentation body text, etc.\n - Substantive explanatory or narrative text, even if short.\n\n2. Rate the coherency of the line on a 0β5 scale.\n - Coherency measures how much the line reads like a complete, meaningful snippet of natural\n language, not how βgoodβ or βusefulβ it is.\n - Use this scale:\n - 0 β Nonsense, random characters, or garbled text.\n - 1 β Extremely fragmented; almost no meaning.\n - 2 β Partially coherent; clearly incomplete or heavily cut off.\n - 3 β Mostly coherent; understandable but may be slightly awkward or incomplete.\n - 4 β Clear and coherent; reads like a normal sentence or phrase.\n - 5 β Fully coherent and well-formed; fluent and natural.\nOutput\n\nRespond only with a single JSON object, no extra text, using this schema:\n\n{\n \"is_clutter\": true or false,\n \"clutter_reason\": \"<short explanation of why it is or isn't clutter>\",\n \"coherency_score\": 0-5,\n \"coherency_reason\": \"<short explanation of the coherency score>\"\n}\n\n- is_clutter must be true if the line is web clutter, otherwise false.\n- clutter_reason should be 1β2 short sentences.\n- coherency_score must be an integer from 0 to 5.\n- coherency_reason should briefly justify the score in 1β2 short sentences.\n- Do not include any commentary outside the JSON.\n\nLine:\n\n0 votes",
"role": "user"
}
] |
[
{
"role": "assistant",
"content": "{\n \"is_clutter\": true,\n \"clutter_reason\": \"This is a vote counter, typical of UI elements used in interactive content.\", \n \"coherency_score\": 5,\n \"coherency_reason\": \"The phrase '0 votes' is a clear, grammatically correct and meaningful statement.\"\n}"
}
] |
0 votes
| 0
| 0.7
|
Qwen/Qwen3-4B-Instruct-2507
| 4,096
|
[
{
"content": "\nYou are a classifier that analyzes a single line of text taken from a web page.\n\nYour task has two parts:\n\n1. Classify whether the line is web clutter or main content.\n - Web clutter includes:\n - Navigation links or menus (e.g., \"Home | About | Contact\")\n - Cookie banners and consent messages\n - Sign-up / login prompts and account bars\n - Ads, promotional banners, calls to action (e.g., \"Buy now\", \"Limited time offer\")\n - Footer boilerplate (e.g., copyright, privacy policy, terms of service)\n - Sidebar widgets and tag clouds\n - Pagination / UI controls (e.g., \"Next page\", \"1 2 3 4\", \"Share on Twitter\")\n - Search bars and filter/sort controls\n - Generic UI labels or button text disconnected from article content\n - And more.\n - Main content includes:\n - Titles of text, attribution of text to users in forums, etc.\n - Sentences that appear to be part of the main article, blog post, product description,\n documentation body text, etc.\n - Substantive explanatory or narrative text, even if short.\n\n2. Rate the coherency of the line on a 0β5 scale.\n - Coherency measures how much the line reads like a complete, meaningful snippet of natural\n language, not how βgoodβ or βusefulβ it is.\n - Use this scale:\n - 0 β Nonsense, random characters, or garbled text.\n - 1 β Extremely fragmented; almost no meaning.\n - 2 β Partially coherent; clearly incomplete or heavily cut off.\n - 3 β Mostly coherent; understandable but may be slightly awkward or incomplete.\n - 4 β Clear and coherent; reads like a normal sentence or phrase.\n - 5 β Fully coherent and well-formed; fluent and natural.\nOutput\n\nRespond only with a single JSON object, no extra text, using this schema:\n\n{\n \"is_clutter\": true or false,\n \"clutter_reason\": \"<short explanation of why it is or isn't clutter>\",\n \"coherency_score\": 0-5,\n \"coherency_reason\": \"<short explanation of the coherency score>\"\n}\n\n- is_clutter must be true if the line is web clutter, otherwise false.\n- clutter_reason should be 1β2 short sentences.\n- coherency_score must be an integer from 0 to 5.\n- coherency_reason should briefly justify the score in 1β2 short sentences.\n- Do not include any commentary outside the JSON.\n\nLine:\n\nPage last updated at 08:48 GMT, Thursday, 24 April 2008 09:48 UK",
"role": "user"
}
] |
[
{
"role": "assistant",
"content": "{\n \"is_clutter\": true,\n \"clutter_reason\": \"This is a timestamp and metadata, typical of footer or page metadata sections.\", \n \"coherency_score\": 5,\n \"coherency_reason\": \"The line is a complete, grammatically correct, and clearly understandable timestamp phrase.\"\n}"
}
] |
Page last updated at 08:48 GMT, Thursday, 24 April 2008 09:48 UK
| 0
| 0.7
|
Qwen/Qwen3-4B-Instruct-2507
| 4,096
|
[
{
"content": "\nYou are a classifier that analyzes a single line of text taken from a web page.\n\nYour task has two parts:\n\n1. Classify whether the line is web clutter or main content.\n - Web clutter includes:\n - Navigation links or menus (e.g., \"Home | About | Contact\")\n - Cookie banners and consent messages\n - Sign-up / login prompts and account bars\n - Ads, promotional banners, calls to action (e.g., \"Buy now\", \"Limited time offer\")\n - Footer boilerplate (e.g., copyright, privacy policy, terms of service)\n - Sidebar widgets and tag clouds\n - Pagination / UI controls (e.g., \"Next page\", \"1 2 3 4\", \"Share on Twitter\")\n - Search bars and filter/sort controls\n - Generic UI labels or button text disconnected from article content\n - And more.\n - Main content includes:\n - Titles of text, attribution of text to users in forums, etc.\n - Sentences that appear to be part of the main article, blog post, product description,\n documentation body text, etc.\n - Substantive explanatory or narrative text, even if short.\n\n2. Rate the coherency of the line on a 0β5 scale.\n - Coherency measures how much the line reads like a complete, meaningful snippet of natural\n language, not how βgoodβ or βusefulβ it is.\n - Use this scale:\n - 0 β Nonsense, random characters, or garbled text.\n - 1 β Extremely fragmented; almost no meaning.\n - 2 β Partially coherent; clearly incomplete or heavily cut off.\n - 3 β Mostly coherent; understandable but may be slightly awkward or incomplete.\n - 4 β Clear and coherent; reads like a normal sentence or phrase.\n - 5 β Fully coherent and well-formed; fluent and natural.\nOutput\n\nRespond only with a single JSON object, no extra text, using this schema:\n\n{\n \"is_clutter\": true or false,\n \"clutter_reason\": \"<short explanation of why it is or isn't clutter>\",\n \"coherency_score\": 0-5,\n \"coherency_reason\": \"<short explanation of the coherency score>\"\n}\n\n- is_clutter must be true if the line is web clutter, otherwise false.\n- clutter_reason should be 1β2 short sentences.\n- coherency_score must be an integer from 0 to 5.\n- coherency_reason should briefly justify the score in 1β2 short sentences.\n- Do not include any commentary outside the JSON.\n\nLine:\n\nCIE-LAB 41.628, 12.193, 21.758",
"role": "user"
}
] |
[
{
"role": "assistant",
"content": "{\n \"is_clutter\": true,\n \"clutter_reason\": \"CIE-LAB values are technical data typically found in image or color analysis sections, often part of metadata or tool outputs rather than main content.\",\n \"coherency_score\": 3,\n \"coherency_reason\": \"The line is technically coherent as a set of numbers, but lacks context or natural language structure, making it feel like a fragmented data point.\"\n}"
}
] |
CIE-LAB 41.628, 12.193, 21.758
| 0
| 0.7
|
Qwen/Qwen3-4B-Instruct-2507
| 4,096
|
[
{
"content": "\nYou are a classifier that analyzes a single line of text taken from a web page.\n\nYour task has two parts:\n\n1. Classify whether the line is web clutter or main content.\n - Web clutter includes:\n - Navigation links or menus (e.g., \"Home | About | Contact\")\n - Cookie banners and consent messages\n - Sign-up / login prompts and account bars\n - Ads, promotional banners, calls to action (e.g., \"Buy now\", \"Limited time offer\")\n - Footer boilerplate (e.g., copyright, privacy policy, terms of service)\n - Sidebar widgets and tag clouds\n - Pagination / UI controls (e.g., \"Next page\", \"1 2 3 4\", \"Share on Twitter\")\n - Search bars and filter/sort controls\n - Generic UI labels or button text disconnected from article content\n - And more.\n - Main content includes:\n - Titles of text, attribution of text to users in forums, etc.\n - Sentences that appear to be part of the main article, blog post, product description,\n documentation body text, etc.\n - Substantive explanatory or narrative text, even if short.\n\n2. Rate the coherency of the line on a 0β5 scale.\n - Coherency measures how much the line reads like a complete, meaningful snippet of natural\n language, not how βgoodβ or βusefulβ it is.\n - Use this scale:\n - 0 β Nonsense, random characters, or garbled text.\n - 1 β Extremely fragmented; almost no meaning.\n - 2 β Partially coherent; clearly incomplete or heavily cut off.\n - 3 β Mostly coherent; understandable but may be slightly awkward or incomplete.\n - 4 β Clear and coherent; reads like a normal sentence or phrase.\n - 5 β Fully coherent and well-formed; fluent and natural.\nOutput\n\nRespond only with a single JSON object, no extra text, using this schema:\n\n{\n \"is_clutter\": true or false,\n \"clutter_reason\": \"<short explanation of why it is or isn't clutter>\",\n \"coherency_score\": 0-5,\n \"coherency_reason\": \"<short explanation of the coherency score>\"\n}\n\n- is_clutter must be true if the line is web clutter, otherwise false.\n- clutter_reason should be 1β2 short sentences.\n- coherency_score must be an integer from 0 to 5.\n- coherency_reason should briefly justify the score in 1β2 short sentences.\n- Do not include any commentary outside the JSON.\n\nLine:\n\nOct 5th, 2011, 10:13 PM",
"role": "user"
}
] |
[
{
"role": "assistant",
"content": "{\n \"is_clutter\": true,\n \"clutter_reason\": \"This is a date and time stamp, commonly found in footers or metadata sections of web pages.\", \n \"coherency_score\": 5,\n \"coherency_reason\": \"The line is a clear, well-formed date and time format, readable and grammatically correct as a standalone phrase.\"\n}"
}
] |
Oct 5th, 2011, 10:13 PM
| 0
| 0.7
|
Qwen/Qwen3-4B-Instruct-2507
| 4,096
|
[
{
"content": "\nYou are a classifier that analyzes a single line of text taken from a web page.\n\nYour task has two parts:\n\n1. Classify whether the line is web clutter or main content.\n - Web clutter includes:\n - Navigation links or menus (e.g., \"Home | About | Contact\")\n - Cookie banners and consent messages\n - Sign-up / login prompts and account bars\n - Ads, promotional banners, calls to action (e.g., \"Buy now\", \"Limited time offer\")\n - Footer boilerplate (e.g., copyright, privacy policy, terms of service)\n - Sidebar widgets and tag clouds\n - Pagination / UI controls (e.g., \"Next page\", \"1 2 3 4\", \"Share on Twitter\")\n - Search bars and filter/sort controls\n - Generic UI labels or button text disconnected from article content\n - And more.\n - Main content includes:\n - Titles of text, attribution of text to users in forums, etc.\n - Sentences that appear to be part of the main article, blog post, product description,\n documentation body text, etc.\n - Substantive explanatory or narrative text, even if short.\n\n2. Rate the coherency of the line on a 0β5 scale.\n - Coherency measures how much the line reads like a complete, meaningful snippet of natural\n language, not how βgoodβ or βusefulβ it is.\n - Use this scale:\n - 0 β Nonsense, random characters, or garbled text.\n - 1 β Extremely fragmented; almost no meaning.\n - 2 β Partially coherent; clearly incomplete or heavily cut off.\n - 3 β Mostly coherent; understandable but may be slightly awkward or incomplete.\n - 4 β Clear and coherent; reads like a normal sentence or phrase.\n - 5 β Fully coherent and well-formed; fluent and natural.\nOutput\n\nRespond only with a single JSON object, no extra text, using this schema:\n\n{\n \"is_clutter\": true or false,\n \"clutter_reason\": \"<short explanation of why it is or isn't clutter>\",\n \"coherency_score\": 0-5,\n \"coherency_reason\": \"<short explanation of the coherency score>\"\n}\n\n- is_clutter must be true if the line is web clutter, otherwise false.\n- clutter_reason should be 1β2 short sentences.\n- coherency_score must be an integer from 0 to 5.\n- coherency_reason should briefly justify the score in 1β2 short sentences.\n- Do not include any commentary outside the JSON.\n\nLine:\n\nοΏ½οΏ½ -\"",
"role": "user"
}
] |
[
{
"role": "assistant",
"content": "{\n \"is_clutter\": true,\n \"clutter_reason\": \"Contains garbled or corrupted characters, typical of encoding issues or web clutter artifacts.\",\n \"coherency_score\": 0,\n \"coherency_reason\": \"Nonsense and random characters with no meaningful structure or language.\"\n}"
}
] |
οΏ½οΏ½ -"
| 0
| 0.7
|
Qwen/Qwen3-4B-Instruct-2507
| 4,096
|
[
{
"content": "\nYou are a classifier that analyzes a single line of text taken from a web page.\n\nYour task has two parts:\n\n1. Classify whether the line is web clutter or main content.\n - Web clutter includes:\n - Navigation links or menus (e.g., \"Home | About | Contact\")\n - Cookie banners and consent messages\n - Sign-up / login prompts and account bars\n - Ads, promotional banners, calls to action (e.g., \"Buy now\", \"Limited time offer\")\n - Footer boilerplate (e.g., copyright, privacy policy, terms of service)\n - Sidebar widgets and tag clouds\n - Pagination / UI controls (e.g., \"Next page\", \"1 2 3 4\", \"Share on Twitter\")\n - Search bars and filter/sort controls\n - Generic UI labels or button text disconnected from article content\n - And more.\n - Main content includes:\n - Titles of text, attribution of text to users in forums, etc.\n - Sentences that appear to be part of the main article, blog post, product description,\n documentation body text, etc.\n - Substantive explanatory or narrative text, even if short.\n\n2. Rate the coherency of the line on a 0β5 scale.\n - Coherency measures how much the line reads like a complete, meaningful snippet of natural\n language, not how βgoodβ or βusefulβ it is.\n - Use this scale:\n - 0 β Nonsense, random characters, or garbled text.\n - 1 β Extremely fragmented; almost no meaning.\n - 2 β Partially coherent; clearly incomplete or heavily cut off.\n - 3 β Mostly coherent; understandable but may be slightly awkward or incomplete.\n - 4 β Clear and coherent; reads like a normal sentence or phrase.\n - 5 β Fully coherent and well-formed; fluent and natural.\nOutput\n\nRespond only with a single JSON object, no extra text, using this schema:\n\n{\n \"is_clutter\": true or false,\n \"clutter_reason\": \"<short explanation of why it is or isn't clutter>\",\n \"coherency_score\": 0-5,\n \"coherency_reason\": \"<short explanation of the coherency score>\"\n}\n\n- is_clutter must be true if the line is web clutter, otherwise false.\n- clutter_reason should be 1β2 short sentences.\n- coherency_score must be an integer from 0 to 5.\n- coherency_reason should briefly justify the score in 1β2 short sentences.\n- Do not include any commentary outside the JSON.\n\nLine:\n\nAriyawansa, HA; Jaklitsch, WM; Voglmayr, H.",
"role": "user"
}
] |
[
{
"role": "assistant",
"content": "{\n \"is_clutter\": false,\n \"clutter_reason\": \"This appears to be a citation or author list, likely part of academic or scientific content.\",\n \"coherency_score\": 5,\n \"coherency_reason\": \"The line is fully coherent and well-formed, following a standard author citation format.\"\n}"
}
] |
Ariyawansa, HA; Jaklitsch, WM; Voglmayr, H.
| 1
| 0.7
|
Qwen/Qwen3-4B-Instruct-2507
| 4,096
|
[
{
"content": "\nYou are a classifier that analyzes a single line of text taken from a web page.\n\nYour task has two parts:\n\n1. Classify whether the line is web clutter or main content.\n - Web clutter includes:\n - Navigation links or menus (e.g., \"Home | About | Contact\")\n - Cookie banners and consent messages\n - Sign-up / login prompts and account bars\n - Ads, promotional banners, calls to action (e.g., \"Buy now\", \"Limited time offer\")\n - Footer boilerplate (e.g., copyright, privacy policy, terms of service)\n - Sidebar widgets and tag clouds\n - Pagination / UI controls (e.g., \"Next page\", \"1 2 3 4\", \"Share on Twitter\")\n - Search bars and filter/sort controls\n - Generic UI labels or button text disconnected from article content\n - And more.\n - Main content includes:\n - Titles of text, attribution of text to users in forums, etc.\n - Sentences that appear to be part of the main article, blog post, product description,\n documentation body text, etc.\n - Substantive explanatory or narrative text, even if short.\n\n2. Rate the coherency of the line on a 0β5 scale.\n - Coherency measures how much the line reads like a complete, meaningful snippet of natural\n language, not how βgoodβ or βusefulβ it is.\n - Use this scale:\n - 0 β Nonsense, random characters, or garbled text.\n - 1 β Extremely fragmented; almost no meaning.\n - 2 β Partially coherent; clearly incomplete or heavily cut off.\n - 3 β Mostly coherent; understandable but may be slightly awkward or incomplete.\n - 4 β Clear and coherent; reads like a normal sentence or phrase.\n - 5 β Fully coherent and well-formed; fluent and natural.\nOutput\n\nRespond only with a single JSON object, no extra text, using this schema:\n\n{\n \"is_clutter\": true or false,\n \"clutter_reason\": \"<short explanation of why it is or isn't clutter>\",\n \"coherency_score\": 0-5,\n \"coherency_reason\": \"<short explanation of the coherency score>\"\n}\n\n- is_clutter must be true if the line is web clutter, otherwise false.\n- clutter_reason should be 1β2 short sentences.\n- coherency_score must be an integer from 0 to 5.\n- coherency_reason should briefly justify the score in 1β2 short sentences.\n- Do not include any commentary outside the JSON.\n\nLine:\n\nGermany was already on heightened alert following a July 19 attack near the Bavarian town of Treuchtlingen, in which a 17-year-old Afghan refugee screaming βAllahu Akbarβ and wielding an axe was able to woundΒ three people and injure another before being shot dead by police.",
"role": "user"
}
] |
[
{
"role": "assistant",
"content": "{\n \"is_clutter\": false,\n \"clutter_reason\": \"This is a substantive narrative sentence describing a real-world event, part of main article content.\",\n \"coherency_score\": 5,\n \"coherency_reason\": \"The sentence is fully coherent, well-formed, and reads naturally as a complete and fluent piece of news reporting.\"\n}"
}
] |
Germany was already on heightened alert following a July 19 attack near the Bavarian town of Treuchtlingen, in which a 17-year-old Afghan refugee screaming βAllahu Akbarβ and wielding an axe was able to woundΒ three people and injure another before being shot dead by police.
| 1
| 0.7
|
Qwen/Qwen3-4B-Instruct-2507
| 4,096
|
[
{
"content": "\nYou are a classifier that analyzes a single line of text taken from a web page.\n\nYour task has two parts:\n\n1. Classify whether the line is web clutter or main content.\n - Web clutter includes:\n - Navigation links or menus (e.g., \"Home | About | Contact\")\n - Cookie banners and consent messages\n - Sign-up / login prompts and account bars\n - Ads, promotional banners, calls to action (e.g., \"Buy now\", \"Limited time offer\")\n - Footer boilerplate (e.g., copyright, privacy policy, terms of service)\n - Sidebar widgets and tag clouds\n - Pagination / UI controls (e.g., \"Next page\", \"1 2 3 4\", \"Share on Twitter\")\n - Search bars and filter/sort controls\n - Generic UI labels or button text disconnected from article content\n - And more.\n - Main content includes:\n - Titles of text, attribution of text to users in forums, etc.\n - Sentences that appear to be part of the main article, blog post, product description,\n documentation body text, etc.\n - Substantive explanatory or narrative text, even if short.\n\n2. Rate the coherency of the line on a 0β5 scale.\n - Coherency measures how much the line reads like a complete, meaningful snippet of natural\n language, not how βgoodβ or βusefulβ it is.\n - Use this scale:\n - 0 β Nonsense, random characters, or garbled text.\n - 1 β Extremely fragmented; almost no meaning.\n - 2 β Partially coherent; clearly incomplete or heavily cut off.\n - 3 β Mostly coherent; understandable but may be slightly awkward or incomplete.\n - 4 β Clear and coherent; reads like a normal sentence or phrase.\n - 5 β Fully coherent and well-formed; fluent and natural.\nOutput\n\nRespond only with a single JSON object, no extra text, using this schema:\n\n{\n \"is_clutter\": true or false,\n \"clutter_reason\": \"<short explanation of why it is or isn't clutter>\",\n \"coherency_score\": 0-5,\n \"coherency_reason\": \"<short explanation of the coherency score>\"\n}\n\n- is_clutter must be true if the line is web clutter, otherwise false.\n- clutter_reason should be 1β2 short sentences.\n- coherency_score must be an integer from 0 to 5.\n- coherency_reason should briefly justify the score in 1β2 short sentences.\n- Do not include any commentary outside the JSON.\n\nLine:\n\n>>115107934>Still sucking Ben mauled penisHAPPY ANNIVERSARY",
"role": "user"
}
] |
[
{
"role": "assistant",
"content": "{\n \"is_clutter\": true,\n \"clutter_reason\": \"Contains explicit and inappropriate content that appears to be spam or user-generated noise rather than meaningful content.\",\n \"coherency_score\": 2,\n \"coherency_reason\": \"Partially coherent but fragmented and semantically jarring; the phrase mixes unrelated elements with explicit language, making it difficult to interpret as a meaningful statement.\"\n}"
}
] |
>>115107934>Still sucking Ben mauled penisHAPPY ANNIVERSARY
| 0
| 0.7
|
Qwen/Qwen3-4B-Instruct-2507
| 4,096
|
[
{
"content": "\nYou are a classifier that analyzes a single line of text taken from a web page.\n\nYour task has two parts:\n\n1. Classify whether the line is web clutter or main content.\n - Web clutter includes:\n - Navigation links or menus (e.g., \"Home | About | Contact\")\n - Cookie banners and consent messages\n - Sign-up / login prompts and account bars\n - Ads, promotional banners, calls to action (e.g., \"Buy now\", \"Limited time offer\")\n - Footer boilerplate (e.g., copyright, privacy policy, terms of service)\n - Sidebar widgets and tag clouds\n - Pagination / UI controls (e.g., \"Next page\", \"1 2 3 4\", \"Share on Twitter\")\n - Search bars and filter/sort controls\n - Generic UI labels or button text disconnected from article content\n - And more.\n - Main content includes:\n - Titles of text, attribution of text to users in forums, etc.\n - Sentences that appear to be part of the main article, blog post, product description,\n documentation body text, etc.\n - Substantive explanatory or narrative text, even if short.\n\n2. Rate the coherency of the line on a 0β5 scale.\n - Coherency measures how much the line reads like a complete, meaningful snippet of natural\n language, not how βgoodβ or βusefulβ it is.\n - Use this scale:\n - 0 β Nonsense, random characters, or garbled text.\n - 1 β Extremely fragmented; almost no meaning.\n - 2 β Partially coherent; clearly incomplete or heavily cut off.\n - 3 β Mostly coherent; understandable but may be slightly awkward or incomplete.\n - 4 β Clear and coherent; reads like a normal sentence or phrase.\n - 5 β Fully coherent and well-formed; fluent and natural.\nOutput\n\nRespond only with a single JSON object, no extra text, using this schema:\n\n{\n \"is_clutter\": true or false,\n \"clutter_reason\": \"<short explanation of why it is or isn't clutter>\",\n \"coherency_score\": 0-5,\n \"coherency_reason\": \"<short explanation of the coherency score>\"\n}\n\n- is_clutter must be true if the line is web clutter, otherwise false.\n- clutter_reason should be 1β2 short sentences.\n- coherency_score must be an integer from 0 to 5.\n- coherency_reason should briefly justify the score in 1β2 short sentences.\n- Do not include any commentary outside the JSON.\n\nLine:\n\nMohjiveΒ ( 2018-09-18 18:36:43 +0200 )edit",
"role": "user"
}
] |
[
{
"role": "assistant",
"content": "{\n \"is_clutter\": true,\n \"clutter_reason\": \"This appears to be a user attribution and timestamp from a forum or wiki interface, typical of navigation or UI metadata.\", \n \"coherency_score\": 4,\n \"coherency_reason\": \"The line is mostly coherent, clearly stating a user and timestamp, though slightly awkward in formatting.\"\n}"
}
] |
MohjiveΒ ( 2018-09-18 18:36:43 +0200 )edit
| 0
| 0.7
|
Qwen/Qwen3-4B-Instruct-2507
| 4,096
|
[
{
"content": "\nYou are a classifier that analyzes a single line of text taken from a web page.\n\nYour task has two parts:\n\n1. Classify whether the line is web clutter or main content.\n - Web clutter includes:\n - Navigation links or menus (e.g., \"Home | About | Contact\")\n - Cookie banners and consent messages\n - Sign-up / login prompts and account bars\n - Ads, promotional banners, calls to action (e.g., \"Buy now\", \"Limited time offer\")\n - Footer boilerplate (e.g., copyright, privacy policy, terms of service)\n - Sidebar widgets and tag clouds\n - Pagination / UI controls (e.g., \"Next page\", \"1 2 3 4\", \"Share on Twitter\")\n - Search bars and filter/sort controls\n - Generic UI labels or button text disconnected from article content\n - And more.\n - Main content includes:\n - Titles of text, attribution of text to users in forums, etc.\n - Sentences that appear to be part of the main article, blog post, product description,\n documentation body text, etc.\n - Substantive explanatory or narrative text, even if short.\n\n2. Rate the coherency of the line on a 0β5 scale.\n - Coherency measures how much the line reads like a complete, meaningful snippet of natural\n language, not how βgoodβ or βusefulβ it is.\n - Use this scale:\n - 0 β Nonsense, random characters, or garbled text.\n - 1 β Extremely fragmented; almost no meaning.\n - 2 β Partially coherent; clearly incomplete or heavily cut off.\n - 3 β Mostly coherent; understandable but may be slightly awkward or incomplete.\n - 4 β Clear and coherent; reads like a normal sentence or phrase.\n - 5 β Fully coherent and well-formed; fluent and natural.\nOutput\n\nRespond only with a single JSON object, no extra text, using this schema:\n\n{\n \"is_clutter\": true or false,\n \"clutter_reason\": \"<short explanation of why it is or isn't clutter>\",\n \"coherency_score\": 0-5,\n \"coherency_reason\": \"<short explanation of the coherency score>\"\n}\n\n- is_clutter must be true if the line is web clutter, otherwise false.\n- clutter_reason should be 1β2 short sentences.\n- coherency_score must be an integer from 0 to 5.\n- coherency_reason should briefly justify the score in 1β2 short sentences.\n- Do not include any commentary outside the JSON.\n\nLine:\n\n min(len(item) for item in args)",
"role": "user"
}
] |
[
{
"role": "assistant",
"content": "{\n \"is_clutter\": true,\n \"clutter_reason\": \"This is a Python code snippet, likely part of a script or function, not natural content or article text.\",\n \"coherency_score\": 2,\n \"coherency_reason\": \"The line is fragmented and syntactically incomplete, lacking clear grammatical structure or natural language flow.\"\n}"
}
] |
min(len(item) for item in args)
| 0
| 0.7
|
Qwen/Qwen3-4B-Instruct-2507
| 4,096
|
[
{
"content": "\nYou are a classifier that analyzes a single line of text taken from a web page.\n\nYour task has two parts:\n\n1. Classify whether the line is web clutter or main content.\n - Web clutter includes:\n - Navigation links or menus (e.g., \"Home | About | Contact\")\n - Cookie banners and consent messages\n - Sign-up / login prompts and account bars\n - Ads, promotional banners, calls to action (e.g., \"Buy now\", \"Limited time offer\")\n - Footer boilerplate (e.g., copyright, privacy policy, terms of service)\n - Sidebar widgets and tag clouds\n - Pagination / UI controls (e.g., \"Next page\", \"1 2 3 4\", \"Share on Twitter\")\n - Search bars and filter/sort controls\n - Generic UI labels or button text disconnected from article content\n - And more.\n - Main content includes:\n - Titles of text, attribution of text to users in forums, etc.\n - Sentences that appear to be part of the main article, blog post, product description,\n documentation body text, etc.\n - Substantive explanatory or narrative text, even if short.\n\n2. Rate the coherency of the line on a 0β5 scale.\n - Coherency measures how much the line reads like a complete, meaningful snippet of natural\n language, not how βgoodβ or βusefulβ it is.\n - Use this scale:\n - 0 β Nonsense, random characters, or garbled text.\n - 1 β Extremely fragmented; almost no meaning.\n - 2 β Partially coherent; clearly incomplete or heavily cut off.\n - 3 β Mostly coherent; understandable but may be slightly awkward or incomplete.\n - 4 β Clear and coherent; reads like a normal sentence or phrase.\n - 5 β Fully coherent and well-formed; fluent and natural.\nOutput\n\nRespond only with a single JSON object, no extra text, using this schema:\n\n{\n \"is_clutter\": true or false,\n \"clutter_reason\": \"<short explanation of why it is or isn't clutter>\",\n \"coherency_score\": 0-5,\n \"coherency_reason\": \"<short explanation of the coherency score>\"\n}\n\n- is_clutter must be true if the line is web clutter, otherwise false.\n- clutter_reason should be 1β2 short sentences.\n- coherency_score must be an integer from 0 to 5.\n- coherency_reason should briefly justify the score in 1β2 short sentences.\n- Do not include any commentary outside the JSON.\n\nLine:\n\nNo. 908225",
"role": "user"
}
] |
[
{
"role": "assistant",
"content": "{\n \"is_clutter\": true,\n \"clutter_reason\": \"This appears to be a numerical identifier, likely a thread or post number in a forum or discussion board.\", \n \"coherency_score\": 3,\n \"coherency_reason\": \"The line is partially coherent as a label or identifier but lacks context or natural language structure.\"\n}"
}
] |
No. 908225
| 0
| 0.7
|
Qwen/Qwen3-4B-Instruct-2507
| 4,096
|
[
{
"content": "\nYou are a classifier that analyzes a single line of text taken from a web page.\n\nYour task has two parts:\n\n1. Classify whether the line is web clutter or main content.\n - Web clutter includes:\n - Navigation links or menus (e.g., \"Home | About | Contact\")\n - Cookie banners and consent messages\n - Sign-up / login prompts and account bars\n - Ads, promotional banners, calls to action (e.g., \"Buy now\", \"Limited time offer\")\n - Footer boilerplate (e.g., copyright, privacy policy, terms of service)\n - Sidebar widgets and tag clouds\n - Pagination / UI controls (e.g., \"Next page\", \"1 2 3 4\", \"Share on Twitter\")\n - Search bars and filter/sort controls\n - Generic UI labels or button text disconnected from article content\n - And more.\n - Main content includes:\n - Titles of text, attribution of text to users in forums, etc.\n - Sentences that appear to be part of the main article, blog post, product description,\n documentation body text, etc.\n - Substantive explanatory or narrative text, even if short.\n\n2. Rate the coherency of the line on a 0β5 scale.\n - Coherency measures how much the line reads like a complete, meaningful snippet of natural\n language, not how βgoodβ or βusefulβ it is.\n - Use this scale:\n - 0 β Nonsense, random characters, or garbled text.\n - 1 β Extremely fragmented; almost no meaning.\n - 2 β Partially coherent; clearly incomplete or heavily cut off.\n - 3 β Mostly coherent; understandable but may be slightly awkward or incomplete.\n - 4 β Clear and coherent; reads like a normal sentence or phrase.\n - 5 β Fully coherent and well-formed; fluent and natural.\nOutput\n\nRespond only with a single JSON object, no extra text, using this schema:\n\n{\n \"is_clutter\": true or false,\n \"clutter_reason\": \"<short explanation of why it is or isn't clutter>\",\n \"coherency_score\": 0-5,\n \"coherency_reason\": \"<short explanation of the coherency score>\"\n}\n\n- is_clutter must be true if the line is web clutter, otherwise false.\n- clutter_reason should be 1β2 short sentences.\n- coherency_score must be an integer from 0 to 5.\n- coherency_reason should briefly justify the score in 1β2 short sentences.\n- Do not include any commentary outside the JSON.\n\nLine:\n\n 32. Just who Iβd go to for advice on vaginal health: an old retired CV surgeon guy. Everything south of the diaphragm is a mystery to him, Iβm sure. I wish I had a dollar for every time Iβve been insulted or patronized by a CV surgeon. Of course when they want a new central line placed in their patient at 2 AM they deign to ask for me.",
"role": "user"
}
] |
[
{
"role": "assistant",
"content": "{\n \"is_clutter\": false,\n \"clutter_reason\": \"The line is part of a narrative discussing personal experience with CV surgeons and vaginal health, which is substantive and contextually relevant to the main content.\",\n \"coherency_score\": 5,\n \"coherency_reason\": \"The sentence is fully coherent, well-formed, and reads naturally as a complete and fluent piece of narrative text.\"\n}"
}
] |
32. Just who Iβd go to for advice on vaginal health: an old retired CV surgeon guy. Everything south of the diaphragm is a mystery to him, Iβm sure. I wish I had a dollar for every time Iβve been insulted or patronized by a CV surgeon. Of course when they want a new central line placed in their patient at 2 AM they deign to ask for me.
| 1
| 0.7
|
Qwen/Qwen3-4B-Instruct-2507
| 4,096
|
[
{
"content": "\nYou are a classifier that analyzes a single line of text taken from a web page.\n\nYour task has two parts:\n\n1. Classify whether the line is web clutter or main content.\n - Web clutter includes:\n - Navigation links or menus (e.g., \"Home | About | Contact\")\n - Cookie banners and consent messages\n - Sign-up / login prompts and account bars\n - Ads, promotional banners, calls to action (e.g., \"Buy now\", \"Limited time offer\")\n - Footer boilerplate (e.g., copyright, privacy policy, terms of service)\n - Sidebar widgets and tag clouds\n - Pagination / UI controls (e.g., \"Next page\", \"1 2 3 4\", \"Share on Twitter\")\n - Search bars and filter/sort controls\n - Generic UI labels or button text disconnected from article content\n - And more.\n - Main content includes:\n - Titles of text, attribution of text to users in forums, etc.\n - Sentences that appear to be part of the main article, blog post, product description,\n documentation body text, etc.\n - Substantive explanatory or narrative text, even if short.\n\n2. Rate the coherency of the line on a 0β5 scale.\n - Coherency measures how much the line reads like a complete, meaningful snippet of natural\n language, not how βgoodβ or βusefulβ it is.\n - Use this scale:\n - 0 β Nonsense, random characters, or garbled text.\n - 1 β Extremely fragmented; almost no meaning.\n - 2 β Partially coherent; clearly incomplete or heavily cut off.\n - 3 β Mostly coherent; understandable but may be slightly awkward or incomplete.\n - 4 β Clear and coherent; reads like a normal sentence or phrase.\n - 5 β Fully coherent and well-formed; fluent and natural.\nOutput\n\nRespond only with a single JSON object, no extra text, using this schema:\n\n{\n \"is_clutter\": true or false,\n \"clutter_reason\": \"<short explanation of why it is or isn't clutter>\",\n \"coherency_score\": 0-5,\n \"coherency_reason\": \"<short explanation of the coherency score>\"\n}\n\n- is_clutter must be true if the line is web clutter, otherwise false.\n- clutter_reason should be 1β2 short sentences.\n- coherency_score must be an integer from 0 to 5.\n- coherency_reason should briefly justify the score in 1β2 short sentences.\n- Do not include any commentary outside the JSON.\n\nLine:\n\nThe crash that almost detonated two nuclear warheads",
"role": "user"
}
] |
[
{
"role": "assistant",
"content": "{\n \"is_clutter\": false,\n \"clutter_reason\": \"This is a substantive narrative statement that appears to be part of a main article or news report.\",\n \"coherency_score\": 5,\n \"coherency_reason\": \"The sentence is fully coherent, grammatically correct, and reads like a natural, fluent snippet of news content.\"\n}"
}
] |
The crash that almost detonated two nuclear warheads
| 1
| 0.7
|
Qwen/Qwen3-4B-Instruct-2507
| 4,096
|
[
{
"content": "\nYou are a classifier that analyzes a single line of text taken from a web page.\n\nYour task has two parts:\n\n1. Classify whether the line is web clutter or main content.\n - Web clutter includes:\n - Navigation links or menus (e.g., \"Home | About | Contact\")\n - Cookie banners and consent messages\n - Sign-up / login prompts and account bars\n - Ads, promotional banners, calls to action (e.g., \"Buy now\", \"Limited time offer\")\n - Footer boilerplate (e.g., copyright, privacy policy, terms of service)\n - Sidebar widgets and tag clouds\n - Pagination / UI controls (e.g., \"Next page\", \"1 2 3 4\", \"Share on Twitter\")\n - Search bars and filter/sort controls\n - Generic UI labels or button text disconnected from article content\n - And more.\n - Main content includes:\n - Titles of text, attribution of text to users in forums, etc.\n - Sentences that appear to be part of the main article, blog post, product description,\n documentation body text, etc.\n - Substantive explanatory or narrative text, even if short.\n\n2. Rate the coherency of the line on a 0β5 scale.\n - Coherency measures how much the line reads like a complete, meaningful snippet of natural\n language, not how βgoodβ or βusefulβ it is.\n - Use this scale:\n - 0 β Nonsense, random characters, or garbled text.\n - 1 β Extremely fragmented; almost no meaning.\n - 2 β Partially coherent; clearly incomplete or heavily cut off.\n - 3 β Mostly coherent; understandable but may be slightly awkward or incomplete.\n - 4 β Clear and coherent; reads like a normal sentence or phrase.\n - 5 β Fully coherent and well-formed; fluent and natural.\nOutput\n\nRespond only with a single JSON object, no extra text, using this schema:\n\n{\n \"is_clutter\": true or false,\n \"clutter_reason\": \"<short explanation of why it is or isn't clutter>\",\n \"coherency_score\": 0-5,\n \"coherency_reason\": \"<short explanation of the coherency score>\"\n}\n\n- is_clutter must be true if the line is web clutter, otherwise false.\n- clutter_reason should be 1β2 short sentences.\n- coherency_score must be an integer from 0 to 5.\n- coherency_reason should briefly justify the score in 1β2 short sentences.\n- Do not include any commentary outside the JSON.\n\nLine:\n\nYou are here",
"role": "user"
}
] |
[
{
"role": "assistant",
"content": "{\n \"is_clutter\": true,\n \"clutter_reason\": \"Β«You are hereΒ» is a navigation breadcrumb label, commonly found in UI navigation systems.\",\n \"coherency_score\": 4,\n \"coherency_reason\": \"The phrase is clear and grammatically complete, reading like a normal navigational instruction.\"\n}"
}
] |
You are here
| 0
| 0.7
|
Qwen/Qwen3-4B-Instruct-2507
| 4,096
|
[
{
"content": "\nYou are a classifier that analyzes a single line of text taken from a web page.\n\nYour task has two parts:\n\n1. Classify whether the line is web clutter or main content.\n - Web clutter includes:\n - Navigation links or menus (e.g., \"Home | About | Contact\")\n - Cookie banners and consent messages\n - Sign-up / login prompts and account bars\n - Ads, promotional banners, calls to action (e.g., \"Buy now\", \"Limited time offer\")\n - Footer boilerplate (e.g., copyright, privacy policy, terms of service)\n - Sidebar widgets and tag clouds\n - Pagination / UI controls (e.g., \"Next page\", \"1 2 3 4\", \"Share on Twitter\")\n - Search bars and filter/sort controls\n - Generic UI labels or button text disconnected from article content\n - And more.\n - Main content includes:\n - Titles of text, attribution of text to users in forums, etc.\n - Sentences that appear to be part of the main article, blog post, product description,\n documentation body text, etc.\n - Substantive explanatory or narrative text, even if short.\n\n2. Rate the coherency of the line on a 0β5 scale.\n - Coherency measures how much the line reads like a complete, meaningful snippet of natural\n language, not how βgoodβ or βusefulβ it is.\n - Use this scale:\n - 0 β Nonsense, random characters, or garbled text.\n - 1 β Extremely fragmented; almost no meaning.\n - 2 β Partially coherent; clearly incomplete or heavily cut off.\n - 3 β Mostly coherent; understandable but may be slightly awkward or incomplete.\n - 4 β Clear and coherent; reads like a normal sentence or phrase.\n - 5 β Fully coherent and well-formed; fluent and natural.\nOutput\n\nRespond only with a single JSON object, no extra text, using this schema:\n\n{\n \"is_clutter\": true or false,\n \"clutter_reason\": \"<short explanation of why it is or isn't clutter>\",\n \"coherency_score\": 0-5,\n \"coherency_reason\": \"<short explanation of the coherency score>\"\n}\n\n- is_clutter must be true if the line is web clutter, otherwise false.\n- clutter_reason should be 1β2 short sentences.\n- coherency_score must be an integer from 0 to 5.\n- coherency_reason should briefly justify the score in 1β2 short sentences.\n- Do not include any commentary outside the JSON.\n\nLine:\n\nTo read the complete poem online, click here.",
"role": "user"
}
] |
[
{
"role": "assistant",
"content": "{\n \"is_clutter\": true,\n \"clutter_reason\": \"This is a call to action directing users to a link, which is typical of promotional or navigational UI elements.\",\n \"coherency_score\": 5,\n \"coherency_reason\": \"The sentence is clear, grammatically correct, and forms a complete, natural phrase.\"\n}"
}
] |
To read the complete poem online, click here.
| 0
| 0.7
|
Qwen/Qwen3-4B-Instruct-2507
| 4,096
|
[
{
"content": "\nYou are a classifier that analyzes a single line of text taken from a web page.\n\nYour task has two parts:\n\n1. Classify whether the line is web clutter or main content.\n - Web clutter includes:\n - Navigation links or menus (e.g., \"Home | About | Contact\")\n - Cookie banners and consent messages\n - Sign-up / login prompts and account bars\n - Ads, promotional banners, calls to action (e.g., \"Buy now\", \"Limited time offer\")\n - Footer boilerplate (e.g., copyright, privacy policy, terms of service)\n - Sidebar widgets and tag clouds\n - Pagination / UI controls (e.g., \"Next page\", \"1 2 3 4\", \"Share on Twitter\")\n - Search bars and filter/sort controls\n - Generic UI labels or button text disconnected from article content\n - And more.\n - Main content includes:\n - Titles of text, attribution of text to users in forums, etc.\n - Sentences that appear to be part of the main article, blog post, product description,\n documentation body text, etc.\n - Substantive explanatory or narrative text, even if short.\n\n2. Rate the coherency of the line on a 0β5 scale.\n - Coherency measures how much the line reads like a complete, meaningful snippet of natural\n language, not how βgoodβ or βusefulβ it is.\n - Use this scale:\n - 0 β Nonsense, random characters, or garbled text.\n - 1 β Extremely fragmented; almost no meaning.\n - 2 β Partially coherent; clearly incomplete or heavily cut off.\n - 3 β Mostly coherent; understandable but may be slightly awkward or incomplete.\n - 4 β Clear and coherent; reads like a normal sentence or phrase.\n - 5 β Fully coherent and well-formed; fluent and natural.\nOutput\n\nRespond only with a single JSON object, no extra text, using this schema:\n\n{\n \"is_clutter\": true or false,\n \"clutter_reason\": \"<short explanation of why it is or isn't clutter>\",\n \"coherency_score\": 0-5,\n \"coherency_reason\": \"<short explanation of the coherency score>\"\n}\n\n- is_clutter must be true if the line is web clutter, otherwise false.\n- clutter_reason should be 1β2 short sentences.\n- coherency_score must be an integer from 0 to 5.\n- coherency_reason should briefly justify the score in 1β2 short sentences.\n- Do not include any commentary outside the JSON.\n\nLine:\n\n β’ Gemini",
"role": "user"
}
] |
[
{
"role": "assistant",
"content": "{\n \"is_clutter\": true,\n \"clutter_reason\": \"This appears to be a bullet point or UI element, likely part of a navigation or feature list.\", \n \"coherency_score\": 3,\n \"coherency_reason\": \"It is a minimal and incomplete phrase, lacking context or full grammatical structure.\"\n}"
}
] |
β’ Gemini
| 0
| 0.7
|
Qwen/Qwen3-4B-Instruct-2507
| 4,096
|
[
{
"content": "\nYou are a classifier that analyzes a single line of text taken from a web page.\n\nYour task has two parts:\n\n1. Classify whether the line is web clutter or main content.\n - Web clutter includes:\n - Navigation links or menus (e.g., \"Home | About | Contact\")\n - Cookie banners and consent messages\n - Sign-up / login prompts and account bars\n - Ads, promotional banners, calls to action (e.g., \"Buy now\", \"Limited time offer\")\n - Footer boilerplate (e.g., copyright, privacy policy, terms of service)\n - Sidebar widgets and tag clouds\n - Pagination / UI controls (e.g., \"Next page\", \"1 2 3 4\", \"Share on Twitter\")\n - Search bars and filter/sort controls\n - Generic UI labels or button text disconnected from article content\n - And more.\n - Main content includes:\n - Titles of text, attribution of text to users in forums, etc.\n - Sentences that appear to be part of the main article, blog post, product description,\n documentation body text, etc.\n - Substantive explanatory or narrative text, even if short.\n\n2. Rate the coherency of the line on a 0β5 scale.\n - Coherency measures how much the line reads like a complete, meaningful snippet of natural\n language, not how βgoodβ or βusefulβ it is.\n - Use this scale:\n - 0 β Nonsense, random characters, or garbled text.\n - 1 β Extremely fragmented; almost no meaning.\n - 2 β Partially coherent; clearly incomplete or heavily cut off.\n - 3 β Mostly coherent; understandable but may be slightly awkward or incomplete.\n - 4 β Clear and coherent; reads like a normal sentence or phrase.\n - 5 β Fully coherent and well-formed; fluent and natural.\nOutput\n\nRespond only with a single JSON object, no extra text, using this schema:\n\n{\n \"is_clutter\": true or false,\n \"clutter_reason\": \"<short explanation of why it is or isn't clutter>\",\n \"coherency_score\": 0-5,\n \"coherency_reason\": \"<short explanation of the coherency score>\"\n}\n\n- is_clutter must be true if the line is web clutter, otherwise false.\n- clutter_reason should be 1β2 short sentences.\n- coherency_score must be an integer from 0 to 5.\n- coherency_reason should briefly justify the score in 1β2 short sentences.\n- Do not include any commentary outside the JSON.\n\nLine:\n\nβ’Β Lou Gering disease",
"role": "user"
}
] |
[
{
"role": "assistant",
"content": "{\n \"is_clutter\": true,\n \"clutter_reason\": \"Lou Gering disease is not a recognized medical condition and appears to be a fabricated or nonsensical term.\",\n \"coherency_score\": 2,\n \"coherency_reason\": \"Partially coherent as it forms a noun phrase, but lacks context or logical structure, suggesting it is invented or erroneous.\"\n}"
}
] |
β’Β Lou Gering disease
| 0
| 0.7
|
Qwen/Qwen3-4B-Instruct-2507
| 4,096
|
[
{
"content": "\nYou are a classifier that analyzes a single line of text taken from a web page.\n\nYour task has two parts:\n\n1. Classify whether the line is web clutter or main content.\n - Web clutter includes:\n - Navigation links or menus (e.g., \"Home | About | Contact\")\n - Cookie banners and consent messages\n - Sign-up / login prompts and account bars\n - Ads, promotional banners, calls to action (e.g., \"Buy now\", \"Limited time offer\")\n - Footer boilerplate (e.g., copyright, privacy policy, terms of service)\n - Sidebar widgets and tag clouds\n - Pagination / UI controls (e.g., \"Next page\", \"1 2 3 4\", \"Share on Twitter\")\n - Search bars and filter/sort controls\n - Generic UI labels or button text disconnected from article content\n - And more.\n - Main content includes:\n - Titles of text, attribution of text to users in forums, etc.\n - Sentences that appear to be part of the main article, blog post, product description,\n documentation body text, etc.\n - Substantive explanatory or narrative text, even if short.\n\n2. Rate the coherency of the line on a 0β5 scale.\n - Coherency measures how much the line reads like a complete, meaningful snippet of natural\n language, not how βgoodβ or βusefulβ it is.\n - Use this scale:\n - 0 β Nonsense, random characters, or garbled text.\n - 1 β Extremely fragmented; almost no meaning.\n - 2 β Partially coherent; clearly incomplete or heavily cut off.\n - 3 β Mostly coherent; understandable but may be slightly awkward or incomplete.\n - 4 β Clear and coherent; reads like a normal sentence or phrase.\n - 5 β Fully coherent and well-formed; fluent and natural.\nOutput\n\nRespond only with a single JSON object, no extra text, using this schema:\n\n{\n \"is_clutter\": true or false,\n \"clutter_reason\": \"<short explanation of why it is or isn't clutter>\",\n \"coherency_score\": 0-5,\n \"coherency_reason\": \"<short explanation of the coherency score>\"\n}\n\n- is_clutter must be true if the line is web clutter, otherwise false.\n- clutter_reason should be 1β2 short sentences.\n- coherency_score must be an integer from 0 to 5.\n- coherency_reason should briefly justify the score in 1β2 short sentences.\n- Do not include any commentary outside the JSON.\n\nLine:\n\nTop NFL Picks Sites",
"role": "user"
}
] |
[
{
"role": "assistant",
"content": "{\n \"is_clutter\": true,\n \"clutter_reason\": \"This is a promotional call to action typical of ad or listing banners targeting sports fans.\",\n \"coherency_score\": 5,\n \"coherency_reason\": \"The phrase is clear, grammatically correct, and reads like a natural, well-formed sentence.\"\n}"
}
] |
Top NFL Picks Sites
| 0
| 0.7
|
Qwen/Qwen3-4B-Instruct-2507
| 4,096
|
[
{
"content": "\nYou are a classifier that analyzes a single line of text taken from a web page.\n\nYour task has two parts:\n\n1. Classify whether the line is web clutter or main content.\n - Web clutter includes:\n - Navigation links or menus (e.g., \"Home | About | Contact\")\n - Cookie banners and consent messages\n - Sign-up / login prompts and account bars\n - Ads, promotional banners, calls to action (e.g., \"Buy now\", \"Limited time offer\")\n - Footer boilerplate (e.g., copyright, privacy policy, terms of service)\n - Sidebar widgets and tag clouds\n - Pagination / UI controls (e.g., \"Next page\", \"1 2 3 4\", \"Share on Twitter\")\n - Search bars and filter/sort controls\n - Generic UI labels or button text disconnected from article content\n - And more.\n - Main content includes:\n - Titles of text, attribution of text to users in forums, etc.\n - Sentences that appear to be part of the main article, blog post, product description,\n documentation body text, etc.\n - Substantive explanatory or narrative text, even if short.\n\n2. Rate the coherency of the line on a 0β5 scale.\n - Coherency measures how much the line reads like a complete, meaningful snippet of natural\n language, not how βgoodβ or βusefulβ it is.\n - Use this scale:\n - 0 β Nonsense, random characters, or garbled text.\n - 1 β Extremely fragmented; almost no meaning.\n - 2 β Partially coherent; clearly incomplete or heavily cut off.\n - 3 β Mostly coherent; understandable but may be slightly awkward or incomplete.\n - 4 β Clear and coherent; reads like a normal sentence or phrase.\n - 5 β Fully coherent and well-formed; fluent and natural.\nOutput\n\nRespond only with a single JSON object, no extra text, using this schema:\n\n{\n \"is_clutter\": true or false,\n \"clutter_reason\": \"<short explanation of why it is or isn't clutter>\",\n \"coherency_score\": 0-5,\n \"coherency_reason\": \"<short explanation of the coherency score>\"\n}\n\n- is_clutter must be true if the line is web clutter, otherwise false.\n- clutter_reason should be 1β2 short sentences.\n- coherency_score must be an integer from 0 to 5.\n- coherency_reason should briefly justify the score in 1β2 short sentences.\n- Do not include any commentary outside the JSON.\n\nLine:\n\nhome | top",
"role": "user"
}
] |
[
{
"role": "assistant",
"content": "{\n \"is_clutter\": true,\n \"clutter_reason\": \"This is a navigation link separator commonly found in headers or footers.\", \n \"coherency_score\": 4,\n \"coherency_reason\": \"The phrase 'home | top' is clear and understandable as a directional link, though slightly informal.\"\n}"
}
] |
home | top
| 0
| 0.7
|
Qwen/Qwen3-4B-Instruct-2507
| 4,096
|
[
{
"content": "\nYou are a classifier that analyzes a single line of text taken from a web page.\n\nYour task has two parts:\n\n1. Classify whether the line is web clutter or main content.\n - Web clutter includes:\n - Navigation links or menus (e.g., \"Home | About | Contact\")\n - Cookie banners and consent messages\n - Sign-up / login prompts and account bars\n - Ads, promotional banners, calls to action (e.g., \"Buy now\", \"Limited time offer\")\n - Footer boilerplate (e.g., copyright, privacy policy, terms of service)\n - Sidebar widgets and tag clouds\n - Pagination / UI controls (e.g., \"Next page\", \"1 2 3 4\", \"Share on Twitter\")\n - Search bars and filter/sort controls\n - Generic UI labels or button text disconnected from article content\n - And more.\n - Main content includes:\n - Titles of text, attribution of text to users in forums, etc.\n - Sentences that appear to be part of the main article, blog post, product description,\n documentation body text, etc.\n - Substantive explanatory or narrative text, even if short.\n\n2. Rate the coherency of the line on a 0β5 scale.\n - Coherency measures how much the line reads like a complete, meaningful snippet of natural\n language, not how βgoodβ or βusefulβ it is.\n - Use this scale:\n - 0 β Nonsense, random characters, or garbled text.\n - 1 β Extremely fragmented; almost no meaning.\n - 2 β Partially coherent; clearly incomplete or heavily cut off.\n - 3 β Mostly coherent; understandable but may be slightly awkward or incomplete.\n - 4 β Clear and coherent; reads like a normal sentence or phrase.\n - 5 β Fully coherent and well-formed; fluent and natural.\nOutput\n\nRespond only with a single JSON object, no extra text, using this schema:\n\n{\n \"is_clutter\": true or false,\n \"clutter_reason\": \"<short explanation of why it is or isn't clutter>\",\n \"coherency_score\": 0-5,\n \"coherency_reason\": \"<short explanation of the coherency score>\"\n}\n\n- is_clutter must be true if the line is web clutter, otherwise false.\n- clutter_reason should be 1β2 short sentences.\n- coherency_score must be an integer from 0 to 5.\n- coherency_reason should briefly justify the score in 1β2 short sentences.\n- Do not include any commentary outside the JSON.\n\nLine:\n\nSatisfied Customers: 55705",
"role": "user"
}
] |
[
{
"role": "assistant",
"content": "{\n \"is_clutter\": true,\n \"clutter_reason\": \"This is a statistic typically found in promotional or footer sections, such as customer satisfaction metrics or sales data displays.\",\n \"coherency_score\": 4,\n \"coherency_reason\": \"The phrase is clear and grammatically complete, stating a measurable fact in a straightforward manner.\"\n}"
}
] |
Satisfied Customers: 55705
| 0
| 0.7
|
Qwen/Qwen3-4B-Instruct-2507
| 4,096
|
[
{
"content": "\nYou are a classifier that analyzes a single line of text taken from a web page.\n\nYour task has two parts:\n\n1. Classify whether the line is web clutter or main content.\n - Web clutter includes:\n - Navigation links or menus (e.g., \"Home | About | Contact\")\n - Cookie banners and consent messages\n - Sign-up / login prompts and account bars\n - Ads, promotional banners, calls to action (e.g., \"Buy now\", \"Limited time offer\")\n - Footer boilerplate (e.g., copyright, privacy policy, terms of service)\n - Sidebar widgets and tag clouds\n - Pagination / UI controls (e.g., \"Next page\", \"1 2 3 4\", \"Share on Twitter\")\n - Search bars and filter/sort controls\n - Generic UI labels or button text disconnected from article content\n - And more.\n - Main content includes:\n - Titles of text, attribution of text to users in forums, etc.\n - Sentences that appear to be part of the main article, blog post, product description,\n documentation body text, etc.\n - Substantive explanatory or narrative text, even if short.\n\n2. Rate the coherency of the line on a 0β5 scale.\n - Coherency measures how much the line reads like a complete, meaningful snippet of natural\n language, not how βgoodβ or βusefulβ it is.\n - Use this scale:\n - 0 β Nonsense, random characters, or garbled text.\n - 1 β Extremely fragmented; almost no meaning.\n - 2 β Partially coherent; clearly incomplete or heavily cut off.\n - 3 β Mostly coherent; understandable but may be slightly awkward or incomplete.\n - 4 β Clear and coherent; reads like a normal sentence or phrase.\n - 5 β Fully coherent and well-formed; fluent and natural.\nOutput\n\nRespond only with a single JSON object, no extra text, using this schema:\n\n{\n \"is_clutter\": true or false,\n \"clutter_reason\": \"<short explanation of why it is or isn't clutter>\",\n \"coherency_score\": 0-5,\n \"coherency_reason\": \"<short explanation of the coherency score>\"\n}\n\n- is_clutter must be true if the line is web clutter, otherwise false.\n- clutter_reason should be 1β2 short sentences.\n- coherency_score must be an integer from 0 to 5.\n- coherency_reason should briefly justify the score in 1β2 short sentences.\n- Do not include any commentary outside the JSON.\n\nLine:\n\n 229. voiceon 11 May 2014 at 4:46 pm",
"role": "user"
}
] |
[
{
"role": "assistant",
"content": "{\n \"is_clutter\": true,\n \"clutter_reason\": \"This appears to be a user attribution line typical of forum or comment sections, indicating a user's identity and timestamp.\",\n \"coherency_score\": 5,\n \"coherency_reason\": \"The line is fully coherent and well-formed, clearly stating a user name, date, and time in a natural format.\"\n}"
}
] |
229. voiceon 11 May 2014 at 4:46 pm
| 0
| 0.7
|
Qwen/Qwen3-4B-Instruct-2507
| 4,096
|
[
{
"content": "\nYou are a classifier that analyzes a single line of text taken from a web page.\n\nYour task has two parts:\n\n1. Classify whether the line is web clutter or main content.\n - Web clutter includes:\n - Navigation links or menus (e.g., \"Home | About | Contact\")\n - Cookie banners and consent messages\n - Sign-up / login prompts and account bars\n - Ads, promotional banners, calls to action (e.g., \"Buy now\", \"Limited time offer\")\n - Footer boilerplate (e.g., copyright, privacy policy, terms of service)\n - Sidebar widgets and tag clouds\n - Pagination / UI controls (e.g., \"Next page\", \"1 2 3 4\", \"Share on Twitter\")\n - Search bars and filter/sort controls\n - Generic UI labels or button text disconnected from article content\n - And more.\n - Main content includes:\n - Titles of text, attribution of text to users in forums, etc.\n - Sentences that appear to be part of the main article, blog post, product description,\n documentation body text, etc.\n - Substantive explanatory or narrative text, even if short.\n\n2. Rate the coherency of the line on a 0β5 scale.\n - Coherency measures how much the line reads like a complete, meaningful snippet of natural\n language, not how βgoodβ or βusefulβ it is.\n - Use this scale:\n - 0 β Nonsense, random characters, or garbled text.\n - 1 β Extremely fragmented; almost no meaning.\n - 2 β Partially coherent; clearly incomplete or heavily cut off.\n - 3 β Mostly coherent; understandable but may be slightly awkward or incomplete.\n - 4 β Clear and coherent; reads like a normal sentence or phrase.\n - 5 β Fully coherent and well-formed; fluent and natural.\nOutput\n\nRespond only with a single JSON object, no extra text, using this schema:\n\n{\n \"is_clutter\": true or false,\n \"clutter_reason\": \"<short explanation of why it is or isn't clutter>\",\n \"coherency_score\": 0-5,\n \"coherency_reason\": \"<short explanation of the coherency score>\"\n}\n\n- is_clutter must be true if the line is web clutter, otherwise false.\n- clutter_reason should be 1β2 short sentences.\n- coherency_score must be an integer from 0 to 5.\n- coherency_reason should briefly justify the score in 1β2 short sentences.\n- Do not include any commentary outside the JSON.\n\nLine:\n\nThere are so many documents and academic books in the world and Sribd has all of them. This, so far, is the biggest online library being accessed by everyone across the globe with a single touch on their phones. There is a better chance that you can find anything you are looking for, be it in a different language or from a particular country. If you are a literature enthusiast, you can learn about different Australia writing styles from this app and how they vary from other countries. This is the best app as it will act as a resource center. This app is also quite helpful to companies that provide writing services Australia to students.Β That is why we can assume that one of the best applications of this top deserves special attention.",
"role": "user"
}
] |
[
{
"role": "assistant",
"content": "{\n \"is_clutter\": false,\n \"clutter_reason\": \"The line describes a product's features and benefits in a narrative form, clearly part of main content related to an app's functionality and value proposition.\",\n \"coherency_score\": 5,\n \"coherency_reason\": \"The text is fully coherent, well-structured, and reads like a natural, fluent paragraph with clear ideas and logical flow.\"\n}"
}
] |
There are so many documents and academic books in the world and Sribd has all of them. This, so far, is the biggest online library being accessed by everyone across the globe with a single touch on their phones. There is a better chance that you can find anything you are looking for, be it in a different language or from a particular country. If you are a literature enthusiast, you can learn about different Australia writing styles from this app and how they vary from other countries. This is the best app as it will act as a resource center. This app is also quite helpful to companies that provide writing services Australia to students.Β That is why we can assume that one of the best applications of this top deserves special attention.
| 1
| 0.7
|
Qwen/Qwen3-4B-Instruct-2507
| 4,096
|
[
{
"content": "\nYou are a classifier that analyzes a single line of text taken from a web page.\n\nYour task has two parts:\n\n1. Classify whether the line is web clutter or main content.\n - Web clutter includes:\n - Navigation links or menus (e.g., \"Home | About | Contact\")\n - Cookie banners and consent messages\n - Sign-up / login prompts and account bars\n - Ads, promotional banners, calls to action (e.g., \"Buy now\", \"Limited time offer\")\n - Footer boilerplate (e.g., copyright, privacy policy, terms of service)\n - Sidebar widgets and tag clouds\n - Pagination / UI controls (e.g., \"Next page\", \"1 2 3 4\", \"Share on Twitter\")\n - Search bars and filter/sort controls\n - Generic UI labels or button text disconnected from article content\n - And more.\n - Main content includes:\n - Titles of text, attribution of text to users in forums, etc.\n - Sentences that appear to be part of the main article, blog post, product description,\n documentation body text, etc.\n - Substantive explanatory or narrative text, even if short.\n\n2. Rate the coherency of the line on a 0β5 scale.\n - Coherency measures how much the line reads like a complete, meaningful snippet of natural\n language, not how βgoodβ or βusefulβ it is.\n - Use this scale:\n - 0 β Nonsense, random characters, or garbled text.\n - 1 β Extremely fragmented; almost no meaning.\n - 2 β Partially coherent; clearly incomplete or heavily cut off.\n - 3 β Mostly coherent; understandable but may be slightly awkward or incomplete.\n - 4 β Clear and coherent; reads like a normal sentence or phrase.\n - 5 β Fully coherent and well-formed; fluent and natural.\nOutput\n\nRespond only with a single JSON object, no extra text, using this schema:\n\n{\n \"is_clutter\": true or false,\n \"clutter_reason\": \"<short explanation of why it is or isn't clutter>\",\n \"coherency_score\": 0-5,\n \"coherency_reason\": \"<short explanation of the coherency score>\"\n}\n\n- is_clutter must be true if the line is web clutter, otherwise false.\n- clutter_reason should be 1β2 short sentences.\n- coherency_score must be an integer from 0 to 5.\n- coherency_reason should briefly justify the score in 1β2 short sentences.\n- Do not include any commentary outside the JSON.\n\nLine:\n\n β’ CF4578",
"role": "user"
}
] |
[
{
"role": "assistant",
"content": "{\n \"is_clutter\": true,\n \"clutter_reason\": \"CF4578 appears to be a code or identifier, likely part of a navigation, search, or UI element rather than substantive content.\",\n \"coherency_score\": 1,\n \"coherency_reason\": \"The line is a meaningless sequence of characters with no grammatical structure or natural language meaning.\"\n}"
}
] |
β’ CF4578
| 0
| 0.7
|
Qwen/Qwen3-4B-Instruct-2507
| 4,096
|
[
{
"content": "\nYou are a classifier that analyzes a single line of text taken from a web page.\n\nYour task has two parts:\n\n1. Classify whether the line is web clutter or main content.\n - Web clutter includes:\n - Navigation links or menus (e.g., \"Home | About | Contact\")\n - Cookie banners and consent messages\n - Sign-up / login prompts and account bars\n - Ads, promotional banners, calls to action (e.g., \"Buy now\", \"Limited time offer\")\n - Footer boilerplate (e.g., copyright, privacy policy, terms of service)\n - Sidebar widgets and tag clouds\n - Pagination / UI controls (e.g., \"Next page\", \"1 2 3 4\", \"Share on Twitter\")\n - Search bars and filter/sort controls\n - Generic UI labels or button text disconnected from article content\n - And more.\n - Main content includes:\n - Titles of text, attribution of text to users in forums, etc.\n - Sentences that appear to be part of the main article, blog post, product description,\n documentation body text, etc.\n - Substantive explanatory or narrative text, even if short.\n\n2. Rate the coherency of the line on a 0β5 scale.\n - Coherency measures how much the line reads like a complete, meaningful snippet of natural\n language, not how βgoodβ or βusefulβ it is.\n - Use this scale:\n - 0 β Nonsense, random characters, or garbled text.\n - 1 β Extremely fragmented; almost no meaning.\n - 2 β Partially coherent; clearly incomplete or heavily cut off.\n - 3 β Mostly coherent; understandable but may be slightly awkward or incomplete.\n - 4 β Clear and coherent; reads like a normal sentence or phrase.\n - 5 β Fully coherent and well-formed; fluent and natural.\nOutput\n\nRespond only with a single JSON object, no extra text, using this schema:\n\n{\n \"is_clutter\": true or false,\n \"clutter_reason\": \"<short explanation of why it is or isn't clutter>\",\n \"coherency_score\": 0-5,\n \"coherency_reason\": \"<short explanation of the coherency score>\"\n}\n\n- is_clutter must be true if the line is web clutter, otherwise false.\n- clutter_reason should be 1β2 short sentences.\n- coherency_score must be an integer from 0 to 5.\n- coherency_reason should briefly justify the score in 1β2 short sentences.\n- Do not include any commentary outside the JSON.\n\nLine:\n\nSome of these operations went relatively well. The UN Compensation Commission in Geneva did about $22 billion worth of business, with no complaints whatsoever. I think it was helped by being in Geneva, away from the politics of New York.",
"role": "user"
}
] |
[
{
"role": "assistant",
"content": "{\n \"is_clutter\": false,\n \"clutter_reason\": \"The line presents a coherent narrative about UN operations and their outcomes, which is part of main content.\",\n \"coherency_score\": 5,\n \"coherency_reason\": \"The sentence is fully coherent, well-formed, and reads naturally as a complete and fluent piece of explanatory text.\"\n}"
}
] |
Some of these operations went relatively well. The UN Compensation Commission in Geneva did about $22 billion worth of business, with no complaints whatsoever. I think it was helped by being in Geneva, away from the politics of New York.
| 1
| 0.7
|
Qwen/Qwen3-4B-Instruct-2507
| 4,096
|
[
{
"content": "\nYou are a classifier that analyzes a single line of text taken from a web page.\n\nYour task has two parts:\n\n1. Classify whether the line is web clutter or main content.\n - Web clutter includes:\n - Navigation links or menus (e.g., \"Home | About | Contact\")\n - Cookie banners and consent messages\n - Sign-up / login prompts and account bars\n - Ads, promotional banners, calls to action (e.g., \"Buy now\", \"Limited time offer\")\n - Footer boilerplate (e.g., copyright, privacy policy, terms of service)\n - Sidebar widgets and tag clouds\n - Pagination / UI controls (e.g., \"Next page\", \"1 2 3 4\", \"Share on Twitter\")\n - Search bars and filter/sort controls\n - Generic UI labels or button text disconnected from article content\n - And more.\n - Main content includes:\n - Titles of text, attribution of text to users in forums, etc.\n - Sentences that appear to be part of the main article, blog post, product description,\n documentation body text, etc.\n - Substantive explanatory or narrative text, even if short.\n\n2. Rate the coherency of the line on a 0β5 scale.\n - Coherency measures how much the line reads like a complete, meaningful snippet of natural\n language, not how βgoodβ or βusefulβ it is.\n - Use this scale:\n - 0 β Nonsense, random characters, or garbled text.\n - 1 β Extremely fragmented; almost no meaning.\n - 2 β Partially coherent; clearly incomplete or heavily cut off.\n - 3 β Mostly coherent; understandable but may be slightly awkward or incomplete.\n - 4 β Clear and coherent; reads like a normal sentence or phrase.\n - 5 β Fully coherent and well-formed; fluent and natural.\nOutput\n\nRespond only with a single JSON object, no extra text, using this schema:\n\n{\n \"is_clutter\": true or false,\n \"clutter_reason\": \"<short explanation of why it is or isn't clutter>\",\n \"coherency_score\": 0-5,\n \"coherency_reason\": \"<short explanation of the coherency score>\"\n}\n\n- is_clutter must be true if the line is web clutter, otherwise false.\n- clutter_reason should be 1β2 short sentences.\n- coherency_score must be an integer from 0 to 5.\n- coherency_reason should briefly justify the score in 1β2 short sentences.\n- Do not include any commentary outside the JSON.\n\nLine:\n\nBookmark and Share",
"role": "user"
}
] |
[
{
"role": "assistant",
"content": "{\n \"is_clutter\": true,\n \"clutter_reason\": \"Bookmark and Share is a generic UI action label typically found in social sharing widgets.\",\n \"coherency_score\": 5,\n \"coherency_reason\": \"The phrase is a complete, well-formed instruction and reads naturally as a standard interface label.\"\n}"
}
] |
Bookmark and Share
| 0
| 0.7
|
Qwen/Qwen3-4B-Instruct-2507
| 4,096
|
[
{
"content": "\nYou are a classifier that analyzes a single line of text taken from a web page.\n\nYour task has two parts:\n\n1. Classify whether the line is web clutter or main content.\n - Web clutter includes:\n - Navigation links or menus (e.g., \"Home | About | Contact\")\n - Cookie banners and consent messages\n - Sign-up / login prompts and account bars\n - Ads, promotional banners, calls to action (e.g., \"Buy now\", \"Limited time offer\")\n - Footer boilerplate (e.g., copyright, privacy policy, terms of service)\n - Sidebar widgets and tag clouds\n - Pagination / UI controls (e.g., \"Next page\", \"1 2 3 4\", \"Share on Twitter\")\n - Search bars and filter/sort controls\n - Generic UI labels or button text disconnected from article content\n - And more.\n - Main content includes:\n - Titles of text, attribution of text to users in forums, etc.\n - Sentences that appear to be part of the main article, blog post, product description,\n documentation body text, etc.\n - Substantive explanatory or narrative text, even if short.\n\n2. Rate the coherency of the line on a 0β5 scale.\n - Coherency measures how much the line reads like a complete, meaningful snippet of natural\n language, not how βgoodβ or βusefulβ it is.\n - Use this scale:\n - 0 β Nonsense, random characters, or garbled text.\n - 1 β Extremely fragmented; almost no meaning.\n - 2 β Partially coherent; clearly incomplete or heavily cut off.\n - 3 β Mostly coherent; understandable but may be slightly awkward or incomplete.\n - 4 β Clear and coherent; reads like a normal sentence or phrase.\n - 5 β Fully coherent and well-formed; fluent and natural.\nOutput\n\nRespond only with a single JSON object, no extra text, using this schema:\n\n{\n \"is_clutter\": true or false,\n \"clutter_reason\": \"<short explanation of why it is or isn't clutter>\",\n \"coherency_score\": 0-5,\n \"coherency_reason\": \"<short explanation of the coherency score>\"\n}\n\n- is_clutter must be true if the line is web clutter, otherwise false.\n- clutter_reason should be 1β2 short sentences.\n- coherency_score must be an integer from 0 to 5.\n- coherency_reason should briefly justify the score in 1β2 short sentences.\n- Do not include any commentary outside the JSON.\n\nLine:\n\n Diphenhydramine Gold Member",
"role": "user"
}
] |
[
{
"role": "assistant",
"content": "{\n \"is_clutter\": true,\n \"clutter_reason\": \"This appears to be a membership or account status label, likely part of a user profile or account bar.\",\n \"coherency_score\": 4,\n \"coherency_reason\": \"The phrase is clear and grammatically complete, though it lacks context. It reads naturally as a label in a UI context.\"\n}"
}
] |
Diphenhydramine Gold Member
| 0
| 0.7
|
Qwen/Qwen3-4B-Instruct-2507
| 4,096
|
[
{
"content": "\nYou are a classifier that analyzes a single line of text taken from a web page.\n\nYour task has two parts:\n\n1. Classify whether the line is web clutter or main content.\n - Web clutter includes:\n - Navigation links or menus (e.g., \"Home | About | Contact\")\n - Cookie banners and consent messages\n - Sign-up / login prompts and account bars\n - Ads, promotional banners, calls to action (e.g., \"Buy now\", \"Limited time offer\")\n - Footer boilerplate (e.g., copyright, privacy policy, terms of service)\n - Sidebar widgets and tag clouds\n - Pagination / UI controls (e.g., \"Next page\", \"1 2 3 4\", \"Share on Twitter\")\n - Search bars and filter/sort controls\n - Generic UI labels or button text disconnected from article content\n - And more.\n - Main content includes:\n - Titles of text, attribution of text to users in forums, etc.\n - Sentences that appear to be part of the main article, blog post, product description,\n documentation body text, etc.\n - Substantive explanatory or narrative text, even if short.\n\n2. Rate the coherency of the line on a 0β5 scale.\n - Coherency measures how much the line reads like a complete, meaningful snippet of natural\n language, not how βgoodβ or βusefulβ it is.\n - Use this scale:\n - 0 β Nonsense, random characters, or garbled text.\n - 1 β Extremely fragmented; almost no meaning.\n - 2 β Partially coherent; clearly incomplete or heavily cut off.\n - 3 β Mostly coherent; understandable but may be slightly awkward or incomplete.\n - 4 β Clear and coherent; reads like a normal sentence or phrase.\n - 5 β Fully coherent and well-formed; fluent and natural.\nOutput\n\nRespond only with a single JSON object, no extra text, using this schema:\n\n{\n \"is_clutter\": true or false,\n \"clutter_reason\": \"<short explanation of why it is or isn't clutter>\",\n \"coherency_score\": 0-5,\n \"coherency_reason\": \"<short explanation of the coherency score>\"\n}\n\n- is_clutter must be true if the line is web clutter, otherwise false.\n- clutter_reason should be 1β2 short sentences.\n- coherency_score must be an integer from 0 to 5.\n- coherency_reason should briefly justify the score in 1β2 short sentences.\n- Do not include any commentary outside the JSON.\n\nLine:\n\nDiscussion in 'Recording' started by Jecjies, Jun 21, 2011.",
"role": "user"
}
] |
[
{
"role": "assistant",
"content": "{\n \"is_clutter\": true,\n \"clutter_reason\": \"This is a forum post header indicating a discussion thread, typical of navigation or metadata in community forums.\",\n \"coherency_score\": 5,\n \"coherency_reason\": \"The line is fully coherent and grammatically correct, forming a natural phrase commonly found in forum interfaces.\"\n}"
}
] |
Discussion in 'Recording' started by Jecjies, Jun 21, 2011.
| 0
| 0.7
|
Qwen/Qwen3-4B-Instruct-2507
| 4,096
|
[
{
"content": "\nYou are a classifier that analyzes a single line of text taken from a web page.\n\nYour task has two parts:\n\n1. Classify whether the line is web clutter or main content.\n - Web clutter includes:\n - Navigation links or menus (e.g., \"Home | About | Contact\")\n - Cookie banners and consent messages\n - Sign-up / login prompts and account bars\n - Ads, promotional banners, calls to action (e.g., \"Buy now\", \"Limited time offer\")\n - Footer boilerplate (e.g., copyright, privacy policy, terms of service)\n - Sidebar widgets and tag clouds\n - Pagination / UI controls (e.g., \"Next page\", \"1 2 3 4\", \"Share on Twitter\")\n - Search bars and filter/sort controls\n - Generic UI labels or button text disconnected from article content\n - And more.\n - Main content includes:\n - Titles of text, attribution of text to users in forums, etc.\n - Sentences that appear to be part of the main article, blog post, product description,\n documentation body text, etc.\n - Substantive explanatory or narrative text, even if short.\n\n2. Rate the coherency of the line on a 0β5 scale.\n - Coherency measures how much the line reads like a complete, meaningful snippet of natural\n language, not how βgoodβ or βusefulβ it is.\n - Use this scale:\n - 0 β Nonsense, random characters, or garbled text.\n - 1 β Extremely fragmented; almost no meaning.\n - 2 β Partially coherent; clearly incomplete or heavily cut off.\n - 3 β Mostly coherent; understandable but may be slightly awkward or incomplete.\n - 4 β Clear and coherent; reads like a normal sentence or phrase.\n - 5 β Fully coherent and well-formed; fluent and natural.\nOutput\n\nRespond only with a single JSON object, no extra text, using this schema:\n\n{\n \"is_clutter\": true or false,\n \"clutter_reason\": \"<short explanation of why it is or isn't clutter>\",\n \"coherency_score\": 0-5,\n \"coherency_reason\": \"<short explanation of the coherency score>\"\n}\n\n- is_clutter must be true if the line is web clutter, otherwise false.\n- clutter_reason should be 1β2 short sentences.\n- coherency_score must be an integer from 0 to 5.\n- coherency_reason should briefly justify the score in 1β2 short sentences.\n- Do not include any commentary outside the JSON.\n\nLine:\n\nThere was a clear and Hemp Cbd Vs Cbd Reddit sweet cbd oil maui voice amazon hemp pain relief cream from the womans mouth Zhulong, where has your eldest brother Zulonghuang gone? The woman cbd oil patch asked with doubts, even with such a Cbd Extract And Dmso look, it was difficult to hide this woman.",
"role": "user"
}
] |
[
{
"role": "assistant",
"content": "{\n \"is_clutter\": false,\n \"clutter_reason\": \"The line contains narrative text that appears to be part of a story or article about cbd oil and related topics.\",\n \"coherency_score\": 3,\n \"coherency_reason\": \"The sentence is mostly coherent but contains disjointed and surreal elements that disrupt natural flow, such as 'Hemp Cbd Vs Cbd Reddit' and 'Zhulong, where has your eldest brother Zulonghuang gone?'\"\n}"
}
] |
There was a clear and Hemp Cbd Vs Cbd Reddit sweet cbd oil maui voice amazon hemp pain relief cream from the womans mouth Zhulong, where has your eldest brother Zulonghuang gone? The woman cbd oil patch asked with doubts, even with such a Cbd Extract And Dmso look, it was difficult to hide this woman.
| 1
| 0.7
|
Qwen/Qwen3-4B-Instruct-2507
| 4,096
|
[
{
"content": "\nYou are a classifier that analyzes a single line of text taken from a web page.\n\nYour task has two parts:\n\n1. Classify whether the line is web clutter or main content.\n - Web clutter includes:\n - Navigation links or menus (e.g., \"Home | About | Contact\")\n - Cookie banners and consent messages\n - Sign-up / login prompts and account bars\n - Ads, promotional banners, calls to action (e.g., \"Buy now\", \"Limited time offer\")\n - Footer boilerplate (e.g., copyright, privacy policy, terms of service)\n - Sidebar widgets and tag clouds\n - Pagination / UI controls (e.g., \"Next page\", \"1 2 3 4\", \"Share on Twitter\")\n - Search bars and filter/sort controls\n - Generic UI labels or button text disconnected from article content\n - And more.\n - Main content includes:\n - Titles of text, attribution of text to users in forums, etc.\n - Sentences that appear to be part of the main article, blog post, product description,\n documentation body text, etc.\n - Substantive explanatory or narrative text, even if short.\n\n2. Rate the coherency of the line on a 0β5 scale.\n - Coherency measures how much the line reads like a complete, meaningful snippet of natural\n language, not how βgoodβ or βusefulβ it is.\n - Use this scale:\n - 0 β Nonsense, random characters, or garbled text.\n - 1 β Extremely fragmented; almost no meaning.\n - 2 β Partially coherent; clearly incomplete or heavily cut off.\n - 3 β Mostly coherent; understandable but may be slightly awkward or incomplete.\n - 4 β Clear and coherent; reads like a normal sentence or phrase.\n - 5 β Fully coherent and well-formed; fluent and natural.\nOutput\n\nRespond only with a single JSON object, no extra text, using this schema:\n\n{\n \"is_clutter\": true or false,\n \"clutter_reason\": \"<short explanation of why it is or isn't clutter>\",\n \"coherency_score\": 0-5,\n \"coherency_reason\": \"<short explanation of the coherency score>\"\n}\n\n- is_clutter must be true if the line is web clutter, otherwise false.\n- clutter_reason should be 1β2 short sentences.\n- coherency_score must be an integer from 0 to 5.\n- coherency_reason should briefly justify the score in 1β2 short sentences.\n- Do not include any commentary outside the JSON.\n\nLine:\n\nIt makes me notice my mixed feelings and thoughts about post-catastrophe aid, with all itβs different views and angles, complications and the diversity of itβs realisation. How much technology and help do the people here need? While it is crucial, how much benefit does short-term aid really leave behind?",
"role": "user"
}
] |
[
{
"role": "assistant",
"content": "{\n \"is_clutter\": false,\n \"clutter_reason\": \"The line presents a substantive narrative reflecting on post-catastrophe aid and human needs.\", \n \"coherency_score\": 5,\n \"coherency_reason\": \"The sentence is fully coherent, well-structured, and flows naturally as a complete thought with clear logical connections.\"\n}"
}
] |
It makes me notice my mixed feelings and thoughts about post-catastrophe aid, with all itβs different views and angles, complications and the diversity of itβs realisation. How much technology and help do the people here need? While it is crucial, how much benefit does short-term aid really leave behind?
| 1
| 0.7
|
Qwen/Qwen3-4B-Instruct-2507
| 4,096
|
[
{
"content": "\nYou are a classifier that analyzes a single line of text taken from a web page.\n\nYour task has two parts:\n\n1. Classify whether the line is web clutter or main content.\n - Web clutter includes:\n - Navigation links or menus (e.g., \"Home | About | Contact\")\n - Cookie banners and consent messages\n - Sign-up / login prompts and account bars\n - Ads, promotional banners, calls to action (e.g., \"Buy now\", \"Limited time offer\")\n - Footer boilerplate (e.g., copyright, privacy policy, terms of service)\n - Sidebar widgets and tag clouds\n - Pagination / UI controls (e.g., \"Next page\", \"1 2 3 4\", \"Share on Twitter\")\n - Search bars and filter/sort controls\n - Generic UI labels or button text disconnected from article content\n - And more.\n - Main content includes:\n - Titles of text, attribution of text to users in forums, etc.\n - Sentences that appear to be part of the main article, blog post, product description,\n documentation body text, etc.\n - Substantive explanatory or narrative text, even if short.\n\n2. Rate the coherency of the line on a 0β5 scale.\n - Coherency measures how much the line reads like a complete, meaningful snippet of natural\n language, not how βgoodβ or βusefulβ it is.\n - Use this scale:\n - 0 β Nonsense, random characters, or garbled text.\n - 1 β Extremely fragmented; almost no meaning.\n - 2 β Partially coherent; clearly incomplete or heavily cut off.\n - 3 β Mostly coherent; understandable but may be slightly awkward or incomplete.\n - 4 β Clear and coherent; reads like a normal sentence or phrase.\n - 5 β Fully coherent and well-formed; fluent and natural.\nOutput\n\nRespond only with a single JSON object, no extra text, using this schema:\n\n{\n \"is_clutter\": true or false,\n \"clutter_reason\": \"<short explanation of why it is or isn't clutter>\",\n \"coherency_score\": 0-5,\n \"coherency_reason\": \"<short explanation of the coherency score>\"\n}\n\n- is_clutter must be true if the line is web clutter, otherwise false.\n- clutter_reason should be 1β2 short sentences.\n- coherency_score must be an integer from 0 to 5.\n- coherency_reason should briefly justify the score in 1β2 short sentences.\n- Do not include any commentary outside the JSON.\n\nLine:\n\nBy Moohbear on 1/20/2012 3:45:46 PM , Rating: 2",
"role": "user"
}
] |
[
{
"role": "assistant",
"content": "{\n \"is_clutter\": true,\n \"clutter_reason\": \"This is a user attribution and rating prompt, typical of forum or comment sections, not main article content.\",\n \"coherency_score\": 5,\n \"coherency_reason\": \"The line is grammatically complete and clearly forms a natural, well-structured phrase.\"\n}"
}
] |
By Moohbear on 1/20/2012 3:45:46 PM , Rating: 2
| 0
| 0.7
|
Qwen/Qwen3-4B-Instruct-2507
| 4,096
|
[
{
"content": "\nYou are a classifier that analyzes a single line of text taken from a web page.\n\nYour task has two parts:\n\n1. Classify whether the line is web clutter or main content.\n - Web clutter includes:\n - Navigation links or menus (e.g., \"Home | About | Contact\")\n - Cookie banners and consent messages\n - Sign-up / login prompts and account bars\n - Ads, promotional banners, calls to action (e.g., \"Buy now\", \"Limited time offer\")\n - Footer boilerplate (e.g., copyright, privacy policy, terms of service)\n - Sidebar widgets and tag clouds\n - Pagination / UI controls (e.g., \"Next page\", \"1 2 3 4\", \"Share on Twitter\")\n - Search bars and filter/sort controls\n - Generic UI labels or button text disconnected from article content\n - And more.\n - Main content includes:\n - Titles of text, attribution of text to users in forums, etc.\n - Sentences that appear to be part of the main article, blog post, product description,\n documentation body text, etc.\n - Substantive explanatory or narrative text, even if short.\n\n2. Rate the coherency of the line on a 0β5 scale.\n - Coherency measures how much the line reads like a complete, meaningful snippet of natural\n language, not how βgoodβ or βusefulβ it is.\n - Use this scale:\n - 0 β Nonsense, random characters, or garbled text.\n - 1 β Extremely fragmented; almost no meaning.\n - 2 β Partially coherent; clearly incomplete or heavily cut off.\n - 3 β Mostly coherent; understandable but may be slightly awkward or incomplete.\n - 4 β Clear and coherent; reads like a normal sentence or phrase.\n - 5 β Fully coherent and well-formed; fluent and natural.\nOutput\n\nRespond only with a single JSON object, no extra text, using this schema:\n\n{\n \"is_clutter\": true or false,\n \"clutter_reason\": \"<short explanation of why it is or isn't clutter>\",\n \"coherency_score\": 0-5,\n \"coherency_reason\": \"<short explanation of the coherency score>\"\n}\n\n- is_clutter must be true if the line is web clutter, otherwise false.\n- clutter_reason should be 1β2 short sentences.\n- coherency_score must be an integer from 0 to 5.\n- coherency_reason should briefly justify the score in 1β2 short sentences.\n- Do not include any commentary outside the JSON.\n\nLine:\n\nChoose one that contains alpha hydroxy acid (AHA) that helps clean the skin without peeling and this acid is gentle on sensitive skin, unlike other exfoliating ingredients such as retinoids.",
"role": "user"
}
] |
[
{
"role": "assistant",
"content": "{\n \"is_clutter\": false,\n \"clutter_reason\": \"This is substantive explanatory text about skin care ingredients and their benefits.\",\n \"coherency_score\": 5,\n \"coherency_reason\": \"The sentence is fully coherent, well-formed, and reads naturally as a complete explanatory statement.\"\n}"
}
] |
Choose one that contains alpha hydroxy acid (AHA) that helps clean the skin without peeling and this acid is gentle on sensitive skin, unlike other exfoliating ingredients such as retinoids.
| 1
| 0.7
|
Qwen/Qwen3-4B-Instruct-2507
| 4,096
|
[
{
"content": "\nYou are a classifier that analyzes a single line of text taken from a web page.\n\nYour task has two parts:\n\n1. Classify whether the line is web clutter or main content.\n - Web clutter includes:\n - Navigation links or menus (e.g., \"Home | About | Contact\")\n - Cookie banners and consent messages\n - Sign-up / login prompts and account bars\n - Ads, promotional banners, calls to action (e.g., \"Buy now\", \"Limited time offer\")\n - Footer boilerplate (e.g., copyright, privacy policy, terms of service)\n - Sidebar widgets and tag clouds\n - Pagination / UI controls (e.g., \"Next page\", \"1 2 3 4\", \"Share on Twitter\")\n - Search bars and filter/sort controls\n - Generic UI labels or button text disconnected from article content\n - And more.\n - Main content includes:\n - Titles of text, attribution of text to users in forums, etc.\n - Sentences that appear to be part of the main article, blog post, product description,\n documentation body text, etc.\n - Substantive explanatory or narrative text, even if short.\n\n2. Rate the coherency of the line on a 0β5 scale.\n - Coherency measures how much the line reads like a complete, meaningful snippet of natural\n language, not how βgoodβ or βusefulβ it is.\n - Use this scale:\n - 0 β Nonsense, random characters, or garbled text.\n - 1 β Extremely fragmented; almost no meaning.\n - 2 β Partially coherent; clearly incomplete or heavily cut off.\n - 3 β Mostly coherent; understandable but may be slightly awkward or incomplete.\n - 4 β Clear and coherent; reads like a normal sentence or phrase.\n - 5 β Fully coherent and well-formed; fluent and natural.\nOutput\n\nRespond only with a single JSON object, no extra text, using this schema:\n\n{\n \"is_clutter\": true or false,\n \"clutter_reason\": \"<short explanation of why it is or isn't clutter>\",\n \"coherency_score\": 0-5,\n \"coherency_reason\": \"<short explanation of the coherency score>\"\n}\n\n- is_clutter must be true if the line is web clutter, otherwise false.\n- clutter_reason should be 1β2 short sentences.\n- coherency_score must be an integer from 0 to 5.\n- coherency_reason should briefly justify the score in 1β2 short sentences.\n- Do not include any commentary outside the JSON.\n\nLine:\n\n 1. +70 β0 docs/CKAN_PostgreSQL_Tests.markdown",
"role": "user"
}
] |
[
{
"role": "assistant",
"content": "{\n \"is_clutter\": true,\n \"clutter_reason\": \"This is a file change log entry typical of version control system output, not part of main content.\", \n \"coherency_score\": 4,\n \"coherency_reason\": \"The line is clear and grammatically structured, reading like a standard Git diff notation, though it is technically a technical metadata line.\"\n}"
}
] |
1. +70 β0 docs/CKAN_PostgreSQL_Tests.markdown
| 0
| 0.7
|
Qwen/Qwen3-4B-Instruct-2507
| 4,096
|
[
{
"content": "\nYou are a classifier that analyzes a single line of text taken from a web page.\n\nYour task has two parts:\n\n1. Classify whether the line is web clutter or main content.\n - Web clutter includes:\n - Navigation links or menus (e.g., \"Home | About | Contact\")\n - Cookie banners and consent messages\n - Sign-up / login prompts and account bars\n - Ads, promotional banners, calls to action (e.g., \"Buy now\", \"Limited time offer\")\n - Footer boilerplate (e.g., copyright, privacy policy, terms of service)\n - Sidebar widgets and tag clouds\n - Pagination / UI controls (e.g., \"Next page\", \"1 2 3 4\", \"Share on Twitter\")\n - Search bars and filter/sort controls\n - Generic UI labels or button text disconnected from article content\n - And more.\n - Main content includes:\n - Titles of text, attribution of text to users in forums, etc.\n - Sentences that appear to be part of the main article, blog post, product description,\n documentation body text, etc.\n - Substantive explanatory or narrative text, even if short.\n\n2. Rate the coherency of the line on a 0β5 scale.\n - Coherency measures how much the line reads like a complete, meaningful snippet of natural\n language, not how βgoodβ or βusefulβ it is.\n - Use this scale:\n - 0 β Nonsense, random characters, or garbled text.\n - 1 β Extremely fragmented; almost no meaning.\n - 2 β Partially coherent; clearly incomplete or heavily cut off.\n - 3 β Mostly coherent; understandable but may be slightly awkward or incomplete.\n - 4 β Clear and coherent; reads like a normal sentence or phrase.\n - 5 β Fully coherent and well-formed; fluent and natural.\nOutput\n\nRespond only with a single JSON object, no extra text, using this schema:\n\n{\n \"is_clutter\": true or false,\n \"clutter_reason\": \"<short explanation of why it is or isn't clutter>\",\n \"coherency_score\": 0-5,\n \"coherency_reason\": \"<short explanation of the coherency score>\"\n}\n\n- is_clutter must be true if the line is web clutter, otherwise false.\n- clutter_reason should be 1β2 short sentences.\n- coherency_score must be an integer from 0 to 5.\n- coherency_reason should briefly justify the score in 1β2 short sentences.\n- Do not include any commentary outside the JSON.\n\nLine:\n\nHeather Moulder is an executive coach, attorney,Β and founder ofΒ Course Correction CoachingΒ who is also a cancer survivor.Β Connect with HeatherΒ for tips and strategies on how she can help you overcome the difficulties of a cancer diagnosis and get through your post-treatment PTSD.",
"role": "user"
}
] |
[
{
"role": "assistant",
"content": "{\n \"is_clutter\": false,\n \"clutter_reason\": \"This is a substantive narrative introducing a professional and providing relevant information about their expertise and services.\",\n \"coherency_score\": 5,\n \"coherency_reason\": \"The sentence is fully coherent, well-formed, and reads naturally as a complete and fluent piece of explanatory text.\"\n}"
}
] |
Heather Moulder is an executive coach, attorney,Β and founder ofΒ Course Correction CoachingΒ who is also a cancer survivor.Β Connect with HeatherΒ for tips and strategies on how she can help you overcome the difficulties of a cancer diagnosis and get through your post-treatment PTSD.
| 1
| 0.7
|
Qwen/Qwen3-4B-Instruct-2507
| 4,096
|
[
{
"content": "\nYou are a classifier that analyzes a single line of text taken from a web page.\n\nYour task has two parts:\n\n1. Classify whether the line is web clutter or main content.\n - Web clutter includes:\n - Navigation links or menus (e.g., \"Home | About | Contact\")\n - Cookie banners and consent messages\n - Sign-up / login prompts and account bars\n - Ads, promotional banners, calls to action (e.g., \"Buy now\", \"Limited time offer\")\n - Footer boilerplate (e.g., copyright, privacy policy, terms of service)\n - Sidebar widgets and tag clouds\n - Pagination / UI controls (e.g., \"Next page\", \"1 2 3 4\", \"Share on Twitter\")\n - Search bars and filter/sort controls\n - Generic UI labels or button text disconnected from article content\n - And more.\n - Main content includes:\n - Titles of text, attribution of text to users in forums, etc.\n - Sentences that appear to be part of the main article, blog post, product description,\n documentation body text, etc.\n - Substantive explanatory or narrative text, even if short.\n\n2. Rate the coherency of the line on a 0β5 scale.\n - Coherency measures how much the line reads like a complete, meaningful snippet of natural\n language, not how βgoodβ or βusefulβ it is.\n - Use this scale:\n - 0 β Nonsense, random characters, or garbled text.\n - 1 β Extremely fragmented; almost no meaning.\n - 2 β Partially coherent; clearly incomplete or heavily cut off.\n - 3 β Mostly coherent; understandable but may be slightly awkward or incomplete.\n - 4 β Clear and coherent; reads like a normal sentence or phrase.\n - 5 β Fully coherent and well-formed; fluent and natural.\nOutput\n\nRespond only with a single JSON object, no extra text, using this schema:\n\n{\n \"is_clutter\": true or false,\n \"clutter_reason\": \"<short explanation of why it is or isn't clutter>\",\n \"coherency_score\": 0-5,\n \"coherency_reason\": \"<short explanation of the coherency score>\"\n}\n\n- is_clutter must be true if the line is web clutter, otherwise false.\n- clutter_reason should be 1β2 short sentences.\n- coherency_score must be an integer from 0 to 5.\n- coherency_reason should briefly justify the score in 1β2 short sentences.\n- Do not include any commentary outside the JSON.\n\nLine:\n\nKunal Sah, a 13-year-old eighth-grader, is an angry speller. He lives with his uncle and aunt at their family-owned Ramada Limited Motel in this tough former railroad town in eastern Utah. Kunal is making himself into a great speller by way of unhappiness and the immense pressure he feels to reunite his family, which was blown apart across two continents when his parents were sent back to India last year after being denied political asylum.",
"role": "user"
}
] |
[
{
"role": "assistant",
"content": "{\n \"is_clutter\": false,\n \"clutter_reason\": \"This is narrative text describing a character and his situation, clearly part of main content.\",\n \"coherency_score\": 5,\n \"coherency_reason\": \"The line is fully coherent, well-formed, and reads like a natural, fluent sentence with clear meaning and flow.\"\n}"
}
] |
Kunal Sah, a 13-year-old eighth-grader, is an angry speller. He lives with his uncle and aunt at their family-owned Ramada Limited Motel in this tough former railroad town in eastern Utah. Kunal is making himself into a great speller by way of unhappiness and the immense pressure he feels to reunite his family, which was blown apart across two continents when his parents were sent back to India last year after being denied political asylum.
| 1
| 0.7
|
Qwen/Qwen3-4B-Instruct-2507
| 4,096
|
[
{
"content": "\nYou are a classifier that analyzes a single line of text taken from a web page.\n\nYour task has two parts:\n\n1. Classify whether the line is web clutter or main content.\n - Web clutter includes:\n - Navigation links or menus (e.g., \"Home | About | Contact\")\n - Cookie banners and consent messages\n - Sign-up / login prompts and account bars\n - Ads, promotional banners, calls to action (e.g., \"Buy now\", \"Limited time offer\")\n - Footer boilerplate (e.g., copyright, privacy policy, terms of service)\n - Sidebar widgets and tag clouds\n - Pagination / UI controls (e.g., \"Next page\", \"1 2 3 4\", \"Share on Twitter\")\n - Search bars and filter/sort controls\n - Generic UI labels or button text disconnected from article content\n - And more.\n - Main content includes:\n - Titles of text, attribution of text to users in forums, etc.\n - Sentences that appear to be part of the main article, blog post, product description,\n documentation body text, etc.\n - Substantive explanatory or narrative text, even if short.\n\n2. Rate the coherency of the line on a 0β5 scale.\n - Coherency measures how much the line reads like a complete, meaningful snippet of natural\n language, not how βgoodβ or βusefulβ it is.\n - Use this scale:\n - 0 β Nonsense, random characters, or garbled text.\n - 1 β Extremely fragmented; almost no meaning.\n - 2 β Partially coherent; clearly incomplete or heavily cut off.\n - 3 β Mostly coherent; understandable but may be slightly awkward or incomplete.\n - 4 β Clear and coherent; reads like a normal sentence or phrase.\n - 5 β Fully coherent and well-formed; fluent and natural.\nOutput\n\nRespond only with a single JSON object, no extra text, using this schema:\n\n{\n \"is_clutter\": true or false,\n \"clutter_reason\": \"<short explanation of why it is or isn't clutter>\",\n \"coherency_score\": 0-5,\n \"coherency_reason\": \"<short explanation of the coherency score>\"\n}\n\n- is_clutter must be true if the line is web clutter, otherwise false.\n- clutter_reason should be 1β2 short sentences.\n- coherency_score must be an integer from 0 to 5.\n- coherency_reason should briefly justify the score in 1β2 short sentences.\n- Do not include any commentary outside the JSON.\n\nLine:\n\nSecurity Concerns",
"role": "user"
}
] |
[
{
"role": "assistant",
"content": "{\n \"is_clutter\": false,\n \"clutter_reason\": \"This is a topic-related phrase that could form part of a main article or blog post on security issues.\",\n \"coherency_score\": 5,\n \"coherency_reason\": \"The phrase is fully coherent, well-formed, and reads naturally as a standalone sentence in a discussion about security.\"\n}"
}
] |
Security Concerns
| 1
| 0.7
|
Qwen/Qwen3-4B-Instruct-2507
| 4,096
|
[
{
"content": "\nYou are a classifier that analyzes a single line of text taken from a web page.\n\nYour task has two parts:\n\n1. Classify whether the line is web clutter or main content.\n - Web clutter includes:\n - Navigation links or menus (e.g., \"Home | About | Contact\")\n - Cookie banners and consent messages\n - Sign-up / login prompts and account bars\n - Ads, promotional banners, calls to action (e.g., \"Buy now\", \"Limited time offer\")\n - Footer boilerplate (e.g., copyright, privacy policy, terms of service)\n - Sidebar widgets and tag clouds\n - Pagination / UI controls (e.g., \"Next page\", \"1 2 3 4\", \"Share on Twitter\")\n - Search bars and filter/sort controls\n - Generic UI labels or button text disconnected from article content\n - And more.\n - Main content includes:\n - Titles of text, attribution of text to users in forums, etc.\n - Sentences that appear to be part of the main article, blog post, product description,\n documentation body text, etc.\n - Substantive explanatory or narrative text, even if short.\n\n2. Rate the coherency of the line on a 0β5 scale.\n - Coherency measures how much the line reads like a complete, meaningful snippet of natural\n language, not how βgoodβ or βusefulβ it is.\n - Use this scale:\n - 0 β Nonsense, random characters, or garbled text.\n - 1 β Extremely fragmented; almost no meaning.\n - 2 β Partially coherent; clearly incomplete or heavily cut off.\n - 3 β Mostly coherent; understandable but may be slightly awkward or incomplete.\n - 4 β Clear and coherent; reads like a normal sentence or phrase.\n - 5 β Fully coherent and well-formed; fluent and natural.\nOutput\n\nRespond only with a single JSON object, no extra text, using this schema:\n\n{\n \"is_clutter\": true or false,\n \"clutter_reason\": \"<short explanation of why it is or isn't clutter>\",\n \"coherency_score\": 0-5,\n \"coherency_reason\": \"<short explanation of the coherency score>\"\n}\n\n- is_clutter must be true if the line is web clutter, otherwise false.\n- clutter_reason should be 1β2 short sentences.\n- coherency_score must be an integer from 0 to 5.\n- coherency_reason should briefly justify the score in 1β2 short sentences.\n- Do not include any commentary outside the JSON.\n\nLine:\n\nThis entry was posted in Apple, Design, Innovation, Leadership, marketing, Strategy, Technology and tagged , , , , , , , , , , , . Bookmark the permalink.",
"role": "user"
}
] |
[
{
"role": "assistant",
"content": "{\n \"is_clutter\": true,\n \"clutter_reason\": \"This is a standard blog post metadata footer, typical of WordPress or similar platforms, containing category tags and permalink information.\",\n \"coherency_score\": 5,\n \"coherency_reason\": \"The line is grammatically correct, clearly structured, and reads like a natural snippet of metadata from a blog post.\"\n}"
}
] |
This entry was posted in Apple, Design, Innovation, Leadership, marketing, Strategy, Technology and tagged , , , , , , , , , , , . Bookmark the permalink.
| 0
| 0.7
|
Qwen/Qwen3-4B-Instruct-2507
| 4,096
|
[
{
"content": "\nYou are a classifier that analyzes a single line of text taken from a web page.\n\nYour task has two parts:\n\n1. Classify whether the line is web clutter or main content.\n - Web clutter includes:\n - Navigation links or menus (e.g., \"Home | About | Contact\")\n - Cookie banners and consent messages\n - Sign-up / login prompts and account bars\n - Ads, promotional banners, calls to action (e.g., \"Buy now\", \"Limited time offer\")\n - Footer boilerplate (e.g., copyright, privacy policy, terms of service)\n - Sidebar widgets and tag clouds\n - Pagination / UI controls (e.g., \"Next page\", \"1 2 3 4\", \"Share on Twitter\")\n - Search bars and filter/sort controls\n - Generic UI labels or button text disconnected from article content\n - And more.\n - Main content includes:\n - Titles of text, attribution of text to users in forums, etc.\n - Sentences that appear to be part of the main article, blog post, product description,\n documentation body text, etc.\n - Substantive explanatory or narrative text, even if short.\n\n2. Rate the coherency of the line on a 0β5 scale.\n - Coherency measures how much the line reads like a complete, meaningful snippet of natural\n language, not how βgoodβ or βusefulβ it is.\n - Use this scale:\n - 0 β Nonsense, random characters, or garbled text.\n - 1 β Extremely fragmented; almost no meaning.\n - 2 β Partially coherent; clearly incomplete or heavily cut off.\n - 3 β Mostly coherent; understandable but may be slightly awkward or incomplete.\n - 4 β Clear and coherent; reads like a normal sentence or phrase.\n - 5 β Fully coherent and well-formed; fluent and natural.\nOutput\n\nRespond only with a single JSON object, no extra text, using this schema:\n\n{\n \"is_clutter\": true or false,\n \"clutter_reason\": \"<short explanation of why it is or isn't clutter>\",\n \"coherency_score\": 0-5,\n \"coherency_reason\": \"<short explanation of the coherency score>\"\n}\n\n- is_clutter must be true if the line is web clutter, otherwise false.\n- clutter_reason should be 1β2 short sentences.\n- coherency_score must be an integer from 0 to 5.\n- coherency_reason should briefly justify the score in 1β2 short sentences.\n- Do not include any commentary outside the JSON.\n\nLine:\n\nMyth two: market forces increase quality",
"role": "user"
}
] |
[
{
"role": "assistant",
"content": "{\n \"is_clutter\": false,\n \"clutter_reason\": \"This is a substantive statement that appears to be part of an article or discussion on market forces and quality.\",\n \"coherency_score\": 5,\n \"coherency_reason\": \"The sentence is fully coherent, well-formed, and reads naturally as a complete, meaningful statement.\"\n}"
}
] |
Myth two: market forces increase quality
| 1
| 0.7
|
Qwen/Qwen3-4B-Instruct-2507
| 4,096
|
[
{
"content": "\nYou are a classifier that analyzes a single line of text taken from a web page.\n\nYour task has two parts:\n\n1. Classify whether the line is web clutter or main content.\n - Web clutter includes:\n - Navigation links or menus (e.g., \"Home | About | Contact\")\n - Cookie banners and consent messages\n - Sign-up / login prompts and account bars\n - Ads, promotional banners, calls to action (e.g., \"Buy now\", \"Limited time offer\")\n - Footer boilerplate (e.g., copyright, privacy policy, terms of service)\n - Sidebar widgets and tag clouds\n - Pagination / UI controls (e.g., \"Next page\", \"1 2 3 4\", \"Share on Twitter\")\n - Search bars and filter/sort controls\n - Generic UI labels or button text disconnected from article content\n - And more.\n - Main content includes:\n - Titles of text, attribution of text to users in forums, etc.\n - Sentences that appear to be part of the main article, blog post, product description,\n documentation body text, etc.\n - Substantive explanatory or narrative text, even if short.\n\n2. Rate the coherency of the line on a 0β5 scale.\n - Coherency measures how much the line reads like a complete, meaningful snippet of natural\n language, not how βgoodβ or βusefulβ it is.\n - Use this scale:\n - 0 β Nonsense, random characters, or garbled text.\n - 1 β Extremely fragmented; almost no meaning.\n - 2 β Partially coherent; clearly incomplete or heavily cut off.\n - 3 β Mostly coherent; understandable but may be slightly awkward or incomplete.\n - 4 β Clear and coherent; reads like a normal sentence or phrase.\n - 5 β Fully coherent and well-formed; fluent and natural.\nOutput\n\nRespond only with a single JSON object, no extra text, using this schema:\n\n{\n \"is_clutter\": true or false,\n \"clutter_reason\": \"<short explanation of why it is or isn't clutter>\",\n \"coherency_score\": 0-5,\n \"coherency_reason\": \"<short explanation of the coherency score>\"\n}\n\n- is_clutter must be true if the line is web clutter, otherwise false.\n- clutter_reason should be 1β2 short sentences.\n- coherency_score must be an integer from 0 to 5.\n- coherency_reason should briefly justify the score in 1β2 short sentences.\n- Do not include any commentary outside the JSON.\n\nLine:\n\nNo Thanks!",
"role": "user"
}
] |
[
{
"role": "assistant",
"content": "{\n \"is_clutter\": true,\n \"clutter_reason\": \"This is a typical cookie consent banner response.\", \n \"coherency_score\": 5,\n \"coherency_reason\": \"The phrase 'No Thanks!' is clear, complete, and reads like a natural response in a user interface context.\"\n}"
}
] |
No Thanks!
| 0
| 0.7
|
Qwen/Qwen3-4B-Instruct-2507
| 4,096
|
[
{
"content": "\nYou are a classifier that analyzes a single line of text taken from a web page.\n\nYour task has two parts:\n\n1. Classify whether the line is web clutter or main content.\n - Web clutter includes:\n - Navigation links or menus (e.g., \"Home | About | Contact\")\n - Cookie banners and consent messages\n - Sign-up / login prompts and account bars\n - Ads, promotional banners, calls to action (e.g., \"Buy now\", \"Limited time offer\")\n - Footer boilerplate (e.g., copyright, privacy policy, terms of service)\n - Sidebar widgets and tag clouds\n - Pagination / UI controls (e.g., \"Next page\", \"1 2 3 4\", \"Share on Twitter\")\n - Search bars and filter/sort controls\n - Generic UI labels or button text disconnected from article content\n - And more.\n - Main content includes:\n - Titles of text, attribution of text to users in forums, etc.\n - Sentences that appear to be part of the main article, blog post, product description,\n documentation body text, etc.\n - Substantive explanatory or narrative text, even if short.\n\n2. Rate the coherency of the line on a 0β5 scale.\n - Coherency measures how much the line reads like a complete, meaningful snippet of natural\n language, not how βgoodβ or βusefulβ it is.\n - Use this scale:\n - 0 β Nonsense, random characters, or garbled text.\n - 1 β Extremely fragmented; almost no meaning.\n - 2 β Partially coherent; clearly incomplete or heavily cut off.\n - 3 β Mostly coherent; understandable but may be slightly awkward or incomplete.\n - 4 β Clear and coherent; reads like a normal sentence or phrase.\n - 5 β Fully coherent and well-formed; fluent and natural.\nOutput\n\nRespond only with a single JSON object, no extra text, using this schema:\n\n{\n \"is_clutter\": true or false,\n \"clutter_reason\": \"<short explanation of why it is or isn't clutter>\",\n \"coherency_score\": 0-5,\n \"coherency_reason\": \"<short explanation of the coherency score>\"\n}\n\n- is_clutter must be true if the line is web clutter, otherwise false.\n- clutter_reason should be 1β2 short sentences.\n- coherency_score must be an integer from 0 to 5.\n- coherency_reason should briefly justify the score in 1β2 short sentences.\n- Do not include any commentary outside the JSON.\n\nLine:\n\nRoshd Islamic Shia Website",
"role": "user"
}
] |
[
{
"role": "assistant",
"content": "{\n \"is_clutter\": false,\n \"clutter_reason\": \"This appears to be a title or heading describing a website, likely part of main content related to religious information.\",\n \"coherency_score\": 5,\n \"coherency_reason\": \"The phrase is fully coherent, well-formed, and reads naturally as a descriptive title.\"\n}"
}
] |
Roshd Islamic Shia Website
| 1
| 0.7
|
Qwen/Qwen3-4B-Instruct-2507
| 4,096
|
[
{
"content": "\nYou are a classifier that analyzes a single line of text taken from a web page.\n\nYour task has two parts:\n\n1. Classify whether the line is web clutter or main content.\n - Web clutter includes:\n - Navigation links or menus (e.g., \"Home | About | Contact\")\n - Cookie banners and consent messages\n - Sign-up / login prompts and account bars\n - Ads, promotional banners, calls to action (e.g., \"Buy now\", \"Limited time offer\")\n - Footer boilerplate (e.g., copyright, privacy policy, terms of service)\n - Sidebar widgets and tag clouds\n - Pagination / UI controls (e.g., \"Next page\", \"1 2 3 4\", \"Share on Twitter\")\n - Search bars and filter/sort controls\n - Generic UI labels or button text disconnected from article content\n - And more.\n - Main content includes:\n - Titles of text, attribution of text to users in forums, etc.\n - Sentences that appear to be part of the main article, blog post, product description,\n documentation body text, etc.\n - Substantive explanatory or narrative text, even if short.\n\n2. Rate the coherency of the line on a 0β5 scale.\n - Coherency measures how much the line reads like a complete, meaningful snippet of natural\n language, not how βgoodβ or βusefulβ it is.\n - Use this scale:\n - 0 β Nonsense, random characters, or garbled text.\n - 1 β Extremely fragmented; almost no meaning.\n - 2 β Partially coherent; clearly incomplete or heavily cut off.\n - 3 β Mostly coherent; understandable but may be slightly awkward or incomplete.\n - 4 β Clear and coherent; reads like a normal sentence or phrase.\n - 5 β Fully coherent and well-formed; fluent and natural.\nOutput\n\nRespond only with a single JSON object, no extra text, using this schema:\n\n{\n \"is_clutter\": true or false,\n \"clutter_reason\": \"<short explanation of why it is or isn't clutter>\",\n \"coherency_score\": 0-5,\n \"coherency_reason\": \"<short explanation of the coherency score>\"\n}\n\n- is_clutter must be true if the line is web clutter, otherwise false.\n- clutter_reason should be 1β2 short sentences.\n- coherency_score must be an integer from 0 to 5.\n- coherency_reason should briefly justify the score in 1β2 short sentences.\n- Do not include any commentary outside the JSON.\n\nLine:\n\nis the keto diet dangerous",
"role": "user"
}
] |
[
{
"role": "assistant",
"content": "{\n \"is_clutter\": false,\n \"clutter_reason\": \"The line is a substantive question about a health topic, likely part of a main article or blog post.\", \n \"coherency_score\": 5,\n \"coherency_reason\": \"The sentence is fully coherent, grammatically correct, and reads like a natural, fluent question.\"\n}"
}
] |
is the keto diet dangerous
| 1
| 0.7
|
Qwen/Qwen3-4B-Instruct-2507
| 4,096
|
[
{
"content": "\nYou are a classifier that analyzes a single line of text taken from a web page.\n\nYour task has two parts:\n\n1. Classify whether the line is web clutter or main content.\n - Web clutter includes:\n - Navigation links or menus (e.g., \"Home | About | Contact\")\n - Cookie banners and consent messages\n - Sign-up / login prompts and account bars\n - Ads, promotional banners, calls to action (e.g., \"Buy now\", \"Limited time offer\")\n - Footer boilerplate (e.g., copyright, privacy policy, terms of service)\n - Sidebar widgets and tag clouds\n - Pagination / UI controls (e.g., \"Next page\", \"1 2 3 4\", \"Share on Twitter\")\n - Search bars and filter/sort controls\n - Generic UI labels or button text disconnected from article content\n - And more.\n - Main content includes:\n - Titles of text, attribution of text to users in forums, etc.\n - Sentences that appear to be part of the main article, blog post, product description,\n documentation body text, etc.\n - Substantive explanatory or narrative text, even if short.\n\n2. Rate the coherency of the line on a 0β5 scale.\n - Coherency measures how much the line reads like a complete, meaningful snippet of natural\n language, not how βgoodβ or βusefulβ it is.\n - Use this scale:\n - 0 β Nonsense, random characters, or garbled text.\n - 1 β Extremely fragmented; almost no meaning.\n - 2 β Partially coherent; clearly incomplete or heavily cut off.\n - 3 β Mostly coherent; understandable but may be slightly awkward or incomplete.\n - 4 β Clear and coherent; reads like a normal sentence or phrase.\n - 5 β Fully coherent and well-formed; fluent and natural.\nOutput\n\nRespond only with a single JSON object, no extra text, using this schema:\n\n{\n \"is_clutter\": true or false,\n \"clutter_reason\": \"<short explanation of why it is or isn't clutter>\",\n \"coherency_score\": 0-5,\n \"coherency_reason\": \"<short explanation of the coherency score>\"\n}\n\n- is_clutter must be true if the line is web clutter, otherwise false.\n- clutter_reason should be 1β2 short sentences.\n- coherency_score must be an integer from 0 to 5.\n- coherency_reason should briefly justify the score in 1β2 short sentences.\n- Do not include any commentary outside the JSON.\n\nLine:\n\nLeave a comment",
"role": "user"
}
] |
[
{
"role": "assistant",
"content": "{\n \"is_clutter\": true,\n \"clutter_reason\": \"This is a generic UI prompt typically found in comment sections or interactive forms.\", \n \"coherency_score\": 5,\n \"coherency_reason\": \"The phrase 'Leave a comment' is a clear, well-formed, and natural instruction in standard English.\"\n}"
}
] |
Leave a comment
| 0
| 0.7
|
Qwen/Qwen3-4B-Instruct-2507
| 4,096
|
[
{
"content": "\nYou are a classifier that analyzes a single line of text taken from a web page.\n\nYour task has two parts:\n\n1. Classify whether the line is web clutter or main content.\n - Web clutter includes:\n - Navigation links or menus (e.g., \"Home | About | Contact\")\n - Cookie banners and consent messages\n - Sign-up / login prompts and account bars\n - Ads, promotional banners, calls to action (e.g., \"Buy now\", \"Limited time offer\")\n - Footer boilerplate (e.g., copyright, privacy policy, terms of service)\n - Sidebar widgets and tag clouds\n - Pagination / UI controls (e.g., \"Next page\", \"1 2 3 4\", \"Share on Twitter\")\n - Search bars and filter/sort controls\n - Generic UI labels or button text disconnected from article content\n - And more.\n - Main content includes:\n - Titles of text, attribution of text to users in forums, etc.\n - Sentences that appear to be part of the main article, blog post, product description,\n documentation body text, etc.\n - Substantive explanatory or narrative text, even if short.\n\n2. Rate the coherency of the line on a 0β5 scale.\n - Coherency measures how much the line reads like a complete, meaningful snippet of natural\n language, not how βgoodβ or βusefulβ it is.\n - Use this scale:\n - 0 β Nonsense, random characters, or garbled text.\n - 1 β Extremely fragmented; almost no meaning.\n - 2 β Partially coherent; clearly incomplete or heavily cut off.\n - 3 β Mostly coherent; understandable but may be slightly awkward or incomplete.\n - 4 β Clear and coherent; reads like a normal sentence or phrase.\n - 5 β Fully coherent and well-formed; fluent and natural.\nOutput\n\nRespond only with a single JSON object, no extra text, using this schema:\n\n{\n \"is_clutter\": true or false,\n \"clutter_reason\": \"<short explanation of why it is or isn't clutter>\",\n \"coherency_score\": 0-5,\n \"coherency_reason\": \"<short explanation of the coherency score>\"\n}\n\n- is_clutter must be true if the line is web clutter, otherwise false.\n- clutter_reason should be 1β2 short sentences.\n- coherency_score must be an integer from 0 to 5.\n- coherency_reason should briefly justify the score in 1β2 short sentences.\n- Do not include any commentary outside the JSON.\n\nLine:\n\nJanella Williams 2 years, 8 months ago",
"role": "user"
}
] |
[
{
"role": "assistant",
"content": "{\n \"is_clutter\": false,\n \"clutter_reason\": \"This appears to be a timestamped user attribution, typical in forum or comment sections as part of content context.\",\n \"coherency_score\": 5,\n \"coherency_reason\": \"The line is fully coherent and well-formed, clearly stating a person and a time period in a natural, grammatical way.\"\n}"
}
] |
Janella Williams 2 years, 8 months ago
| 1
| 0.7
|
Qwen/Qwen3-4B-Instruct-2507
| 4,096
|
[
{
"content": "\nYou are a classifier that analyzes a single line of text taken from a web page.\n\nYour task has two parts:\n\n1. Classify whether the line is web clutter or main content.\n - Web clutter includes:\n - Navigation links or menus (e.g., \"Home | About | Contact\")\n - Cookie banners and consent messages\n - Sign-up / login prompts and account bars\n - Ads, promotional banners, calls to action (e.g., \"Buy now\", \"Limited time offer\")\n - Footer boilerplate (e.g., copyright, privacy policy, terms of service)\n - Sidebar widgets and tag clouds\n - Pagination / UI controls (e.g., \"Next page\", \"1 2 3 4\", \"Share on Twitter\")\n - Search bars and filter/sort controls\n - Generic UI labels or button text disconnected from article content\n - And more.\n - Main content includes:\n - Titles of text, attribution of text to users in forums, etc.\n - Sentences that appear to be part of the main article, blog post, product description,\n documentation body text, etc.\n - Substantive explanatory or narrative text, even if short.\n\n2. Rate the coherency of the line on a 0β5 scale.\n - Coherency measures how much the line reads like a complete, meaningful snippet of natural\n language, not how βgoodβ or βusefulβ it is.\n - Use this scale:\n - 0 β Nonsense, random characters, or garbled text.\n - 1 β Extremely fragmented; almost no meaning.\n - 2 β Partially coherent; clearly incomplete or heavily cut off.\n - 3 β Mostly coherent; understandable but may be slightly awkward or incomplete.\n - 4 β Clear and coherent; reads like a normal sentence or phrase.\n - 5 β Fully coherent and well-formed; fluent and natural.\nOutput\n\nRespond only with a single JSON object, no extra text, using this schema:\n\n{\n \"is_clutter\": true or false,\n \"clutter_reason\": \"<short explanation of why it is or isn't clutter>\",\n \"coherency_score\": 0-5,\n \"coherency_reason\": \"<short explanation of the coherency score>\"\n}\n\n- is_clutter must be true if the line is web clutter, otherwise false.\n- clutter_reason should be 1β2 short sentences.\n- coherency_score must be an integer from 0 to 5.\n- coherency_reason should briefly justify the score in 1β2 short sentences.\n- Do not include any commentary outside the JSON.\n\nLine:\n\nClass Project Ideas. Project Proposal and Specification. Design Plan. Design Checkpoints and Documentation.",
"role": "user"
}
] |
[
{
"role": "assistant",
"content": "{\n \"is_clutter\": true,\n \"clutter_reason\": \"These are structured headings typically found in project management or academic navigation menus, not substantive article content.\",\n \"coherency_score\": 5,\n \"coherency_reason\": \"Each line is a clear, grammatically correct phrase that reads like a natural heading in a structured document or outline.\"\n}"
}
] |
Class Project Ideas. Project Proposal and Specification. Design Plan. Design Checkpoints and Documentation.
| 0
| 0.7
|
Qwen/Qwen3-4B-Instruct-2507
| 4,096
|
[
{
"content": "\nYou are a classifier that analyzes a single line of text taken from a web page.\n\nYour task has two parts:\n\n1. Classify whether the line is web clutter or main content.\n - Web clutter includes:\n - Navigation links or menus (e.g., \"Home | About | Contact\")\n - Cookie banners and consent messages\n - Sign-up / login prompts and account bars\n - Ads, promotional banners, calls to action (e.g., \"Buy now\", \"Limited time offer\")\n - Footer boilerplate (e.g., copyright, privacy policy, terms of service)\n - Sidebar widgets and tag clouds\n - Pagination / UI controls (e.g., \"Next page\", \"1 2 3 4\", \"Share on Twitter\")\n - Search bars and filter/sort controls\n - Generic UI labels or button text disconnected from article content\n - And more.\n - Main content includes:\n - Titles of text, attribution of text to users in forums, etc.\n - Sentences that appear to be part of the main article, blog post, product description,\n documentation body text, etc.\n - Substantive explanatory or narrative text, even if short.\n\n2. Rate the coherency of the line on a 0β5 scale.\n - Coherency measures how much the line reads like a complete, meaningful snippet of natural\n language, not how βgoodβ or βusefulβ it is.\n - Use this scale:\n - 0 β Nonsense, random characters, or garbled text.\n - 1 β Extremely fragmented; almost no meaning.\n - 2 β Partially coherent; clearly incomplete or heavily cut off.\n - 3 β Mostly coherent; understandable but may be slightly awkward or incomplete.\n - 4 β Clear and coherent; reads like a normal sentence or phrase.\n - 5 β Fully coherent and well-formed; fluent and natural.\nOutput\n\nRespond only with a single JSON object, no extra text, using this schema:\n\n{\n \"is_clutter\": true or false,\n \"clutter_reason\": \"<short explanation of why it is or isn't clutter>\",\n \"coherency_score\": 0-5,\n \"coherency_reason\": \"<short explanation of the coherency score>\"\n}\n\n- is_clutter must be true if the line is web clutter, otherwise false.\n- clutter_reason should be 1β2 short sentences.\n- coherency_score must be an integer from 0 to 5.\n- coherency_reason should briefly justify the score in 1β2 short sentences.\n- Do not include any commentary outside the JSON.\n\nLine:\n\nAnonymous said...",
"role": "user"
}
] |
[
{
"role": "assistant",
"content": "{\n \"is_clutter\": true,\n \"clutter_reason\": \"This is a user attribution label typical of forum or comment sections, common in web clutter.\",\n \"coherency_score\": 5,\n \"coherency_reason\": \"The phrase is fully coherent and grammatically correct, reading naturally as a user attribution in a discussion context.\"\n}"
}
] |
Anonymous said...
| 0
| 0.7
|
Qwen/Qwen3-4B-Instruct-2507
| 4,096
|
[
{
"content": "\nYou are a classifier that analyzes a single line of text taken from a web page.\n\nYour task has two parts:\n\n1. Classify whether the line is web clutter or main content.\n - Web clutter includes:\n - Navigation links or menus (e.g., \"Home | About | Contact\")\n - Cookie banners and consent messages\n - Sign-up / login prompts and account bars\n - Ads, promotional banners, calls to action (e.g., \"Buy now\", \"Limited time offer\")\n - Footer boilerplate (e.g., copyright, privacy policy, terms of service)\n - Sidebar widgets and tag clouds\n - Pagination / UI controls (e.g., \"Next page\", \"1 2 3 4\", \"Share on Twitter\")\n - Search bars and filter/sort controls\n - Generic UI labels or button text disconnected from article content\n - And more.\n - Main content includes:\n - Titles of text, attribution of text to users in forums, etc.\n - Sentences that appear to be part of the main article, blog post, product description,\n documentation body text, etc.\n - Substantive explanatory or narrative text, even if short.\n\n2. Rate the coherency of the line on a 0β5 scale.\n - Coherency measures how much the line reads like a complete, meaningful snippet of natural\n language, not how βgoodβ or βusefulβ it is.\n - Use this scale:\n - 0 β Nonsense, random characters, or garbled text.\n - 1 β Extremely fragmented; almost no meaning.\n - 2 β Partially coherent; clearly incomplete or heavily cut off.\n - 3 β Mostly coherent; understandable but may be slightly awkward or incomplete.\n - 4 β Clear and coherent; reads like a normal sentence or phrase.\n - 5 β Fully coherent and well-formed; fluent and natural.\nOutput\n\nRespond only with a single JSON object, no extra text, using this schema:\n\n{\n \"is_clutter\": true or false,\n \"clutter_reason\": \"<short explanation of why it is or isn't clutter>\",\n \"coherency_score\": 0-5,\n \"coherency_reason\": \"<short explanation of the coherency score>\"\n}\n\n- is_clutter must be true if the line is web clutter, otherwise false.\n- clutter_reason should be 1β2 short sentences.\n- coherency_score must be an integer from 0 to 5.\n- coherency_reason should briefly justify the score in 1β2 short sentences.\n- Do not include any commentary outside the JSON.\n\nLine:\n\nBut commuting miles? Where is it ordained that living an hour from work and chugging back and forth in a single-occupant vehicle is legitimate? Sorry... commuting doesn't compute. I'm not saying we have to radically reinvent our society, but we do need to consider all the alternatives to long commutes in cars: living closer to our jobs; telecommuting; mass transit; AND cycle-commuting. And while I'm not saying everyone who commutes by car has to stop doing it, I am saying that anyone who is doing it has no room to criticize anyone else for taking up space on the roads...especially cyclists.",
"role": "user"
}
] |
[
{
"role": "assistant",
"content": "{\n \"is_clutter\": false,\n \"clutter_reason\": \"The line presents a substantive argument about commuting and transportation alternatives, clearly part of a main article or blog post.\",\n \"coherency_score\": 5,\n \"coherency_reason\": \"The sentence structure is fluent, logically connected, and reads like a natural, well-formed paragraph from a thoughtful essay.\"\n}"
}
] |
But commuting miles? Where is it ordained that living an hour from work and chugging back and forth in a single-occupant vehicle is legitimate? Sorry... commuting doesn't compute. I'm not saying we have to radically reinvent our society, but we do need to consider all the alternatives to long commutes in cars: living closer to our jobs; telecommuting; mass transit; AND cycle-commuting. And while I'm not saying everyone who commutes by car has to stop doing it, I am saying that anyone who is doing it has no room to criticize anyone else for taking up space on the roads...especially cyclists.
| 1
| 0.7
|
Qwen/Qwen3-4B-Instruct-2507
| 4,096
|
End of preview. Expand
in Data Studio
No dataset card yet
- Downloads last month
- -