v2 update new

app.py

@@ -52,11 +52,10 @@ with gr.Blocks() as block:
                     label="Maximum number of parameters (B)",
                 )
 
-            task_choices = [col for col in COLUMN_NAMES if col not in BASE_COLS]
             with gr.Row():
                 tasks_select = gr.CheckboxGroup(
-                    choices=task_choices,
-                    value=task_choices,
+                    choices=TASKS_V1 + TASKS_V2,
+                    value=TASKS_V1,
                     label="Select tasks to Display",
                     elem_id="tasks-select"
                 )

results.jsonl

@@ -1,30 +1,30 @@
-{"Models":"B3","Model Size(B)":8.29,"Data Source":"Self-Reported","Overall":72.0,"Classification":70.0,"VQA":66.5,"Retrieval":74.1,"Grounding":84.6,"URL":"https:\/\/huggingface.co\/raghavlite\/B3_Qwen2_7B"}
-{"Models":"CLIP-FT","Model Size(B)":0.428,"Data Source":"TIGER-Lab","Overall":45.4,"Classification":55.2,"VQA":19.7,"Retrieval":53.2,"Grounding":62.2,"URL":"https:\/\/doi.org\/10.48550\/arXiv.2103.00020"}
-{"Models":"LLaVE-0.5B","Model Size(B)":0.894,"Data Source":"Self-Reported","Overall":59.1,"Classification":57.4,"VQA":50.3,"Retrieval":59.8,"Grounding":82.9,"URL":"https:\/\/huggingface.co\/zhibinlan\/LLaVE-0.5B"}
-{"Models":"LLaVE-2B","Model Size(B)":1.95,"Data Source":"Self-Reported","Overall":65.2,"Classification":62.1,"VQA":60.2,"Retrieval":65.2,"Grounding":84.9,"URL":"https:\/\/huggingface.co\/zhibinlan\/LLaVE-2B"}
-{"Models":"LLaVE-7B","Model Size(B)":8.03,"Data Source":"Self-Reported","Overall":70.3,"Classification":65.7,"VQA":65.4,"Retrieval":70.9,"Grounding":91.9,"URL":"https:\/\/huggingface.co\/zhibinlan\/LLaVE-7B"}
-{"Models":"MM-Embed","Model Size(B)":8.18,"Data Source":"Self-Reported","Overall":50.0,"Classification":48.1,"VQA":32.3,"Retrieval":63.8,"Grounding":57.8,"URL":"https:\/\/huggingface.co\/nvidia\/MM-Embed"}
-{"Models":"MMRet-MLLM (FT)","Model Size(B)":7.57,"Data Source":"Self-Reported","Overall":64.1,"Classification":56.0,"VQA":57.4,"Retrieval":69.9,"Grounding":83.6,"URL":"https:\/\/huggingface.co\/JUNJIE99\/MMRet-large"}
-{"Models":"MMRet-MLLM (LLaVA-1.6)","Model Size(B)":7.57,"Data Source":"Self-Reported","Overall":44.0,"Classification":47.2,"VQA":18.4,"Retrieval":56.5,"Grounding":62.2,"URL":"https:\/\/huggingface.co\/JUNJIE99\/MMRet-large"}
-{"Models":"Magiclens","Model Size(B)":0.428,"Data Source":"TIGER-Lab","Overall":27.8,"Classification":38.8,"VQA":8.3,"Retrieval":35.4,"Grounding":26.0,"URL":"https:\/\/github.com\/google-deepmind\/magiclens"}
-{"Models":"OpenCLIP-FT","Model Size(B)":0.428,"Data Source":"TIGER-Lab","Overall":47.2,"Classification":56.0,"VQA":21.9,"Retrieval":55.4,"Grounding":64.1,"URL":"https:\/\/doi.org\/10.48550\/arXiv.2212.07143"}
-{"Models":"QQMM-embed","Model Size(B)":8.297,"Data Source":"Self-Reported","Overall":72.175,"Classification":70.07,"VQA":69.52,"Retrieval":71.175,"Grounding":87.075,"URL":"https:\/\/github.com\/QQ-MM\/QQMM-embed"}
-{"Models":"UniIR (BLIP_FF)","Model Size(B)":0.247,"Data Source":"TIGER-Lab","Overall":42.8,"Classification":42.1,"VQA":15.0,"Retrieval":60.1,"Grounding":62.2,"URL":"https:\/\/huggingface.co\/TIGER-Lab\/UniIR"}
-{"Models":"UniIR (CLIP_SF)","Model Size(B)":0.428,"Data Source":"TIGER-Lab","Overall":44.7,"Classification":44.3,"VQA":16.2,"Retrieval":61.8,"Grounding":65.3,"URL":"https:\/\/huggingface.co\/TIGER-Lab\/UniIR"}
-{"Models":"UniME(LLaVA-1.6-7B-LoRA-LowRes)","Model Size(B)":7.57,"Data Source":"Self-Reported","Overall":66.6,"Classification":60.6,"VQA":52.9,"Retrieval":67.9,"Grounding":85.1,"URL":"https:\/\/huggingface.co\/DeepGlint-AI\/UniME-LLaVA-1.6-7B"}
-{"Models":"UniME(LLaVA-OneVision-7B-LoRA-Res336)","Model Size(B)":8.03,"Data Source":"Self-Reported","Overall":70.7,"Classification":66.8,"VQA":66.6,"Retrieval":70.5,"Grounding":90.9,"URL":"https:\/\/huggingface.co\/DeepGlint-AI\/UniME-LLaVA-OneVision-7B"}
-{"Models":"UniME(Phi-3.5-V-LoRA)","Model Size(B)":4.2,"Data Source":"Self-Reported","Overall":64.2,"Classification":54.8,"VQA":55.9,"Retrieval":64.5,"Grounding":81.8,"URL":"https:\/\/huggingface.co\/DeepGlint-AI\/UniME-Phi3.5-V-4.2B"}
-{"Models":"VLM2Vec (LLaVA-1.6-LoRA-HighRes)","Model Size(B)":7.57,"Data Source":"TIGER-Lab","Overall":62.9,"Classification":61.2,"VQA":49.9,"Retrieval":67.4,"Grounding":86.1,"URL":"https://huggingface.co/TIGER-Lab/VLM2Vec-LLaVa-Next"}
-{"Models":"VLM2Vec (LLaVA-1.6-LoRA-LowRes)","Model Size(B)":7.57,"Data Source":"TIGER-Lab","Overall":55.0,"Classification":54.7,"VQA":50.3,"Retrieval":56.2,"Grounding":64.0,"URL":"https://huggingface.co/TIGER-Lab/VLM2Vec-LLaVa-Next"}
-{"Models":"VLM2Vec (Phi-3.5-V-FT)","Model Size(B)":4.15,"Data Source":"TIGER-Lab","Overall":55.9,"Classification":52.8,"VQA":50.3,"Retrieval":57.8,"Grounding":72.3,"URL":"https:\/\/huggingface.co\/TIGER-Lab\/VLM2Vec-Full"}
-{"Models":"VLM2Vec (Phi-3.5-V-LoRA)","Model Size(B)":4.15,"Data Source":"TIGER-Lab","Overall":60.1,"Classification":54.8,"VQA":54.9,"Retrieval":62.3,"Grounding":79.5,"URL":"https:\/\/huggingface.co\/TIGER-Lab\/VLM2Vec-Full"}
-{"Models":"VLM2Vec (Qwen2-VL-2B-LoRA-HighRes)","Model Size(B)":2.21,"Data Source":"TIGER-Lab","Overall":59.3,"Classification":59.0,"VQA":49.4,"Retrieval":65.4,"Grounding":73.4,"URL":"https:\/\/huggingface.co\/TIGER-Lab\/VLM2Vec-Qwen2VL-2B"}
-{"Models":"VLM2Vec (Qwen2-VL-7B-LoRA-HighRes)","Model Size(B)":8.29,"Data Source":"TIGER-Lab","Overall":65.8,"Classification":62.6,"VQA":57.8,"Retrieval":69.9,"Grounding":81.7,"URL":"https:\/\/huggingface.co\/TIGER-Lab\/VLM2Vec-Qwen2VL-7B"}
-{"Models":"blip2-opt-2.7b","Model Size(B)":3.74,"Data Source":"TIGER-Lab","Overall":25.2,"Classification":27.0,"VQA":4.2,"Retrieval":33.9,"Grounding":47.0,"URL":"https:\/\/huggingface.co\/Salesforce\/blip2-opt-2.7b"}
-{"Models":"clip-vit-large-patch14","Model Size(B)":0.428,"Data Source":"TIGER-Lab","Overall":37.8,"Classification":42.8,"VQA":9.1,"Retrieval":53.0,"Grounding":51.8,"URL":"https:\/\/huggingface.co\/openai\/clip-vit-large-patch14"}
-{"Models":"e5-v","Model Size(B)":8.36,"Data Source":"TIGER-Lab","Overall":13.3,"Classification":21.8,"VQA":4.9,"Retrieval":11.5,"Grounding":19.0,"URL":"https:\/\/huggingface.co\/royokong\/e5-v"}
-{"Models":"gme-Qwen2-VL-2B-Instruct","Model Size(B)":2.21,"Data Source":"Self-Reported","Overall":55.8,"Classification":56.9,"VQA":41.2,"Retrieval":67.8,"Grounding":53.4,"URL":"https:\/\/huggingface.co\/Alibaba-NLP\/gme-Qwen2-VL-2B-Instruct"}
-{"Models":"mmE5 (w\/ 560K synthetic data)","Model Size(B)":10.6,"Data Source":"Self-Reported","Overall":58.6,"Classification":60.6,"VQA":55.7,"Retrieval":54.7,"Grounding":72.4,"URL":"https:\/\/huggingface.co\/intfloat\/mmE5-mllama-11b-instruct"}
-{"Models":"mmE5-mllama-11b-instruct","Model Size(B)":10.6,"Data Source":"Self-Reported","Overall":69.8,"Classification":67.6,"VQA":62.6,"Retrieval":71.0,"Grounding":89.6,"URL":"https:\/\/huggingface.co\/intfloat\/mmE5-mllama-11b-instruct"}
-{"Models":"open_clip-ViT-L\/14","Model Size(B)":0.428,"Data Source":"TIGER-Lab","Overall":39.7,"Classification":47.8,"VQA":10.9,"Retrieval":52.3,"Grounding":53.3,"URL":"https:\/\/github.com\/mlfoundations\/open_clip"}
-{"Models":"siglip-base-patch16-224","Model Size(B)":0.203,"Data Source":"TIGER-Lab","Overall":34.8,"Classification":40.3,"VQA":8.4,"Retrieval":31.6,"Grounding":59.5,"URL":"https:\/\/huggingface.co\/google\/siglip-base-patch16-224"}
+{"Models":"B3","Model Size(B)":8.29,"Data Source":"Self-Reported","V2-Overall":null,"V1-Overall":72.0,"I-CLS":70.0,"I-QA":66.5,"I-RET":74.1,"I-VG":84.6,"V-CLS":null,"V-QA":null,"V-RET":null,"V-MRET":null,"VisDoc":null,"URL":"https:\/\/huggingface.co\/raghavlite\/B3_Qwen2_7B"}
+{"Models":"CLIP-FT","Model Size(B)":0.428,"Data Source":"TIGER-Lab","V2-Overall":null,"V1-Overall":45.4,"I-CLS":55.2,"I-QA":19.7,"I-RET":53.2,"I-VG":62.2,"V-CLS":null,"V-QA":null,"V-RET":null,"V-MRET":null,"VisDoc":null,"URL":"https:\/\/doi.org\/10.48550\/arXiv.2103.00020"}
+{"Models":"LLaVE-0.5B","Model Size(B)":0.894,"Data Source":"Self-Reported","V2-Overall":null,"V1-Overall":59.1,"I-CLS":57.4,"I-QA":50.3,"I-RET":59.8,"I-VG":82.9,"V-CLS":null,"V-QA":null,"V-RET":null,"V-MRET":null,"VisDoc":null,"URL":"https:\/\/huggingface.co\/zhibinlan\/LLaVE-0.5B"}
+{"Models":"LLaVE-2B","Model Size(B)":1.95,"Data Source":"Self-Reported","V2-Overall":null,"V1-Overall":65.2,"I-CLS":62.1,"I-QA":60.2,"I-RET":65.2,"I-VG":84.9,"V-CLS":null,"V-QA":null,"V-RET":null,"V-MRET":null,"VisDoc":null,"URL":"https:\/\/huggingface.co\/zhibinlan\/LLaVE-2B"}
+{"Models":"LLaVE-7B","Model Size(B)":8.03,"Data Source":"Self-Reported","V2-Overall":null,"V1-Overall":70.3,"I-CLS":65.7,"I-QA":65.4,"I-RET":70.9,"I-VG":91.9,"V-CLS":null,"V-QA":null,"V-RET":null,"V-MRET":null,"VisDoc":null,"URL":"https:\/\/huggingface.co\/zhibinlan\/LLaVE-7B"}
+{"Models":"MM-Embed","Model Size(B)":8.18,"Data Source":"Self-Reported","V2-Overall":null,"V1-Overall":50.0,"I-CLS":48.1,"I-QA":32.3,"I-RET":63.8,"I-VG":57.8,"V-CLS":null,"V-QA":null,"V-RET":null,"V-MRET":null,"VisDoc":null,"URL":"https:\/\/huggingface.co\/nvidia\/MM-Embed"}
+{"Models":"MMRet-MLLM (FT)","Model Size(B)":7.57,"Data Source":"Self-Reported","V2-Overall":null,"V1-Overall":64.1,"I-CLS":56.0,"I-QA":57.4,"I-RET":69.9,"I-VG":83.6,"V-CLS":null,"V-QA":null,"V-RET":null,"V-MRET":null,"VisDoc":null,"URL":"https:\/\/huggingface.co\/JUNJIE99\/MMRet-large"}
+{"Models":"MMRet-MLLM (LLaVA-1.6)","Model Size(B)":7.57,"Data Source":"Self-Reported","V2-Overall":null,"V1-Overall":44.0,"I-CLS":47.2,"I-QA":18.4,"I-RET":56.5,"I-VG":62.2,"V-CLS":null,"V-QA":null,"V-RET":null,"V-MRET":null,"VisDoc":null,"URL":"https:\/\/huggingface.co\/JUNJIE99\/MMRet-large"}
+{"Models":"Magiclens","Model Size(B)":0.428,"Data Source":"TIGER-Lab","V2-Overall":null,"V1-Overall":27.8,"I-CLS":38.8,"I-QA":8.3,"I-RET":35.4,"I-VG":26.0,"V-CLS":null,"V-QA":null,"V-RET":null,"V-MRET":null,"VisDoc":null,"URL":"https:\/\/github.com\/google-deepmind\/magiclens"}
+{"Models":"OpenCLIP-FT","Model Size(B)":0.428,"Data Source":"TIGER-Lab","V2-Overall":null,"V1-Overall":47.2,"I-CLS":56.0,"I-QA":21.9,"I-RET":55.4,"I-VG":64.1,"V-CLS":null,"V-QA":null,"V-RET":null,"V-MRET":null,"VisDoc":null,"URL":"https:\/\/doi.org\/10.48550\/arXiv.2212.07143"}
+{"Models":"QQMM-embed","Model Size(B)":8.297,"Data Source":"Self-Reported","V2-Overall":null,"V1-Overall":72.175,"I-CLS":70.07,"I-QA":69.52,"I-RET":71.175,"I-VG":87.075,"V-CLS":null,"V-QA":null,"V-RET":null,"V-MRET":null,"VisDoc":null,"URL":"https:\/\/github.com\/QQ-MM\/QQMM-embed"}
+{"Models":"UniIR (BLIP_FF)","Model Size(B)":0.247,"Data Source":"TIGER-Lab","V2-Overall":null,"V1-Overall":42.8,"I-CLS":42.1,"I-QA":15.0,"I-RET":60.1,"I-VG":62.2,"V-CLS":null,"V-QA":null,"V-RET":null,"V-MRET":null,"VisDoc":null,"URL":"https:\/\/huggingface.co\/TIGER-Lab\/UniIR"}
+{"Models":"UniIR (CLIP_SF)","Model Size(B)":0.428,"Data Source":"TIGER-Lab","V2-Overall":null,"V1-Overall":44.7,"I-CLS":44.3,"I-QA":16.2,"I-RET":61.8,"I-VG":65.3,"V-CLS":null,"V-QA":null,"V-RET":null,"V-MRET":null,"VisDoc":null,"URL":"https:\/\/huggingface.co\/TIGER-Lab\/UniIR"}
+{"Models":"UniME(LLaVA-1.6-7B-LoRA-LowRes)","Model Size(B)":7.57,"Data Source":"Self-Reported","V2-Overall":null,"V1-Overall":66.6,"I-CLS":60.6,"I-QA":52.9,"I-RET":67.9,"I-VG":85.1,"V-CLS":null,"V-QA":null,"V-RET":null,"V-MRET":null,"VisDoc":null,"URL":"https:\/\/huggingface.co\/DeepGlint-AI\/UniME-LLaVA-1.6-7B"}
+{"Models":"UniME(LLaVA-OneVision-7B-LoRA-Res336)","Model Size(B)":8.03,"Data Source":"Self-Reported","V2-Overall":null,"V1-Overall":70.7,"I-CLS":66.8,"I-QA":66.6,"I-RET":70.5,"I-VG":90.9,"V-CLS":null,"V-QA":null,"V-RET":null,"V-MRET":null,"VisDoc":null,"URL":"https:\/\/huggingface.co\/DeepGlint-AI\/UniME-LLaVA-OneVision-7B"}
+{"Models":"UniME(Phi-3.5-V-LoRA)","Model Size(B)":4.2,"Data Source":"Self-Reported","V2-Overall":null,"V1-Overall":64.2,"I-CLS":54.8,"I-QA":55.9,"I-RET":64.5,"I-VG":81.8,"V-CLS":null,"V-QA":null,"V-RET":null,"V-MRET":null,"VisDoc":null,"URL":"https:\/\/huggingface.co\/DeepGlint-AI\/UniME-Phi3.5-V-4.2B"}
+{"Models":"VLM2Vec (LLaVA-1.6-LoRA-HighRes)","Model Size(B)":7.57,"Data Source":"TIGER-Lab","V2-Overall":null,"V1-Overall":62.9,"I-CLS":61.2,"I-QA":49.9,"I-RET":67.4,"I-VG":86.1,"V-CLS":null,"V-QA":null,"V-RET":null,"V-MRET":null,"VisDoc":null,"URL":"https:\/\/huggingface.co\/TIGER-Lab\/VLM2Vec-LLaVa-Next"}
+{"Models":"VLM2Vec (LLaVA-1.6-LoRA-LowRes)","Model Size(B)":7.57,"Data Source":"TIGER-Lab","V2-Overall":null,"V1-Overall":55.0,"I-CLS":54.7,"I-QA":50.3,"I-RET":56.2,"I-VG":64.0,"V-CLS":null,"V-QA":null,"V-RET":null,"V-MRET":null,"VisDoc":null,"URL":"https:\/\/huggingface.co\/TIGER-Lab\/VLM2Vec-LLaVa-Next"}
+{"Models":"VLM2Vec (Phi-3.5-V-FT)","Model Size(B)":4.15,"Data Source":"TIGER-Lab","V2-Overall":null,"V1-Overall":55.9,"I-CLS":52.8,"I-QA":50.3,"I-RET":57.8,"I-VG":72.3,"V-CLS":null,"V-QA":null,"V-RET":null,"V-MRET":null,"VisDoc":null,"URL":"https:\/\/huggingface.co\/TIGER-Lab\/VLM2Vec-Full"}
+{"Models":"VLM2Vec (Phi-3.5-V-LoRA)","Model Size(B)":4.15,"Data Source":"TIGER-Lab","V2-Overall":null,"V1-Overall":60.1,"I-CLS":54.8,"I-QA":54.9,"I-RET":62.3,"I-VG":79.5,"V-CLS":null,"V-QA":null,"V-RET":null,"V-MRET":null,"VisDoc":null,"URL":"https:\/\/huggingface.co\/TIGER-Lab\/VLM2Vec-Full"}
+{"Models":"VLM2Vec (Qwen2-VL-2B-LoRA-HighRes)","Model Size(B)":2.21,"Data Source":"TIGER-Lab","V2-Overall":null,"V1-Overall":59.3,"I-CLS":59.0,"I-QA":49.4,"I-RET":65.4,"I-VG":73.4,"V-CLS":null,"V-QA":null,"V-RET":null,"V-MRET":null,"VisDoc":null,"URL":"https:\/\/huggingface.co\/TIGER-Lab\/VLM2Vec-Qwen2VL-2B"}
+{"Models":"VLM2Vec (Qwen2-VL-7B-LoRA-HighRes)","Model Size(B)":8.29,"Data Source":"TIGER-Lab","V2-Overall":null,"V1-Overall":65.8,"I-CLS":62.6,"I-QA":57.8,"I-RET":69.9,"I-VG":81.7,"V-CLS":null,"V-QA":null,"V-RET":null,"V-MRET":null,"VisDoc":null,"URL":"https:\/\/huggingface.co\/TIGER-Lab\/VLM2Vec-Qwen2VL-7B"}
+{"Models":"blip2-opt-2.7b","Model Size(B)":3.74,"Data Source":"TIGER-Lab","V2-Overall":null,"V1-Overall":25.2,"I-CLS":27.0,"I-QA":4.2,"I-RET":33.9,"I-VG":47.0,"V-CLS":null,"V-QA":null,"V-RET":null,"V-MRET":null,"VisDoc":null,"URL":"https:\/\/huggingface.co\/Salesforce\/blip2-opt-2.7b"}
+{"Models":"clip-vit-large-patch14","Model Size(B)":0.428,"Data Source":"TIGER-Lab","V2-Overall":null,"V1-Overall":37.8,"I-CLS":42.8,"I-QA":9.1,"I-RET":53.0,"I-VG":51.8,"V-CLS":null,"V-QA":null,"V-RET":null,"V-MRET":null,"VisDoc":null,"URL":"https:\/\/huggingface.co\/openai\/clip-vit-large-patch14"}
+{"Models":"e5-v","Model Size(B)":8.36,"Data Source":"TIGER-Lab","V2-Overall":null,"V1-Overall":13.3,"I-CLS":21.8,"I-QA":4.9,"I-RET":11.5,"I-VG":19.0,"V-CLS":null,"V-QA":null,"V-RET":null,"V-MRET":null,"VisDoc":null,"URL":"https:\/\/huggingface.co\/royokong\/e5-v"}
+{"Models":"gme-Qwen2-VL-2B-Instruct","Model Size(B)":2.21,"Data Source":"Self-Reported","V2-Overall":null,"V1-Overall":55.8,"I-CLS":56.9,"I-QA":41.2,"I-RET":67.8,"I-VG":53.4,"V-CLS":null,"V-QA":null,"V-RET":null,"V-MRET":null,"VisDoc":null,"URL":"https:\/\/huggingface.co\/Alibaba-NLP\/gme-Qwen2-VL-2B-Instruct"}
+{"Models":"mmE5 (w\/ 560K synthetic data)","Model Size(B)":10.6,"Data Source":"Self-Reported","V2-Overall":null,"V1-Overall":58.6,"I-CLS":60.6,"I-QA":55.7,"I-RET":54.7,"I-VG":72.4,"V-CLS":null,"V-QA":null,"V-RET":null,"V-MRET":null,"VisDoc":null,"URL":"https:\/\/huggingface.co\/intfloat\/mmE5-mllama-11b-instruct"}
+{"Models":"mmE5-mllama-11b-instruct","Model Size(B)":10.6,"Data Source":"Self-Reported","V2-Overall":null,"V1-Overall":69.8,"I-CLS":67.6,"I-QA":62.6,"I-RET":71.0,"I-VG":89.6,"V-CLS":null,"V-QA":null,"V-RET":null,"V-MRET":null,"VisDoc":null,"URL":"https:\/\/huggingface.co\/intfloat\/mmE5-mllama-11b-instruct"}
+{"Models":"open_clip-ViT-L\/14","Model Size(B)":0.428,"Data Source":"TIGER-Lab","V2-Overall":null,"V1-Overall":39.7,"I-CLS":47.8,"I-QA":10.9,"I-RET":52.3,"I-VG":53.3,"V-CLS":null,"V-QA":null,"V-RET":null,"V-MRET":null,"VisDoc":null,"URL":"https:\/\/github.com\/mlfoundations\/open_clip"}
+{"Models":"siglip-base-patch16-224","Model Size(B)":0.203,"Data Source":"TIGER-Lab","V2-Overall":null,"V1-Overall":34.8,"I-CLS":40.3,"I-QA":8.4,"I-RET":31.6,"I-VG":59.5,"V-CLS":null,"V-QA":null,"V-RET":null,"V-MRET":null,"VisDoc":null,"URL":"https:\/\/huggingface.co\/google\/siglip-base-patch16-224"}

utils.py

@@ -10,24 +10,13 @@ from huggingface_hub import Repository
 
 HF_TOKEN = os.environ.get("HF_TOKEN")
 
-TASKS = ["Classification", "VQA", "Retrieval", "Grounding"]
+BASE_COLS = ["Rank", "Models", "Model Size(B)", "Data Source"]
+TASKS_V1 = ["V1-Overall", "I-CLS", "I-QA", "I-RET", "I-VG"]
+TASKS_V2 = ["V2-Overall", "V-CLS", "V-QA", "V-RET", "V-MRET", "VisDoc"]
+COLUMN_NAMES = BASE_COLS + TASKS_V1 + TASKS_V2
 
-MODEL_INFO = [
-    "Rank", "Models", "Model Size(B)", "Data Source",
-    "Overall",
-    "Classification", "VQA", "Retrieval", "Grounding"
-]
-
-BASE_COLS = [col for col in MODEL_INFO if col not in TASKS]
-
-DATA_TITLE_TYPE = ['number', 'markdown', 'str', 'markdown', 'number', 'number', 'number', 'number', 'number', 'number', 'number']
-
-SUBMISSION_NAME = "MMEB"
-SUBMISSION_URL = os.path.join("https://huggingface.co/spaces/TIGER-Lab/", SUBMISSION_NAME)
-FILE_NAME = "results.csv"
-CSV_DIR = "results.csv"
-
-COLUMN_NAMES = MODEL_INFO
+DATA_TITLE_TYPE = ['number', 'markdown', 'str', 'markdown'] + \
+                    ['number'] * (len(TASKS_V1) + len(TASKS_V2))
 
 LEADERBOARD_INTRODUCTION = """
 # MMEB Leaderboard
@@ -49,25 +38,7 @@ TABLE_INTRODUCTION = """"""
 
 LEADERBOARD_INFO = """
 ## Dataset Summary
-MMEB is organized into four primary meta-task categories:
-- **Classification**: This category comprises 5 in-distribution and 5 out-of-distribution datasets. Queries
-consist of instructions and images, optionally accompanied by related text. Targets are class labels,
-and the number of class labels corresponds to the number of classes in the dataset. \n
-        - IND: ImageNet-1k, N24News, HatefulMemes, VOC2007, SUN397 \n
-        - OOD: Place365, ImageNet-A, ImageNet-R, ObjectNet, Country-211 \n
-- **Visual Question Answering**: This category includes 6 in-distribution and 4 out-of-distribution
-datasets. The query consists of an instruction, an image, and a piece of text as the question, while
-the target is the answer. Each query has 1,000 target candidates: 1 ground truth and 999 distractors. \n
-        - IND: OK-VQA, A-OKVQA, DocVQA, InfographicVQA, ChartQA, Visual7W \n
-        - OOD: ScienceQA, VizWiz, GQA, TextVQA \n
-- **Information Retrieval**: This category contains 8 in-distribution and 4 out-of-distribution datasets.
-Both the query and target sides can involve a combination of text, images, and instructions. Similar
-to the VQA task, each query has 1,000 candidates, with 1 ground truth and 999 distractors. \n
-        - IND: VisDial, CIRR, VisualNews_t2i, VisualNews_i2t, MSCOCO_t2i, MSCOCO_i2t, NIGHTS, WebQA \n
-        - OOD: OVEN, FashionIQ, EDIS, Wiki-SS-NQ \n
-- **Visual Grounding**: This category includes 1 in-distribution and 3 out-of-distribution datasets, which are adapted from object detection tasks. Queries consist of an instruction, an image, and text referring to a specific region or object within the image. The target may include a cropped image of the object or text describing the same region. Each query includes 1,000 candidates: 1 ground truth and 999 distractors. These distractors may include hard negatives from the same object class, other objects in the image, or random objects from different images. \n
-        - IND: MSCOCO \n
-        - OOD: Visual7W-Pointing, RefCOCO, RefCOCO-Matching \n
+<img width="900" alt="abs" src="overview.png">
 """
 
 CITATION_BUTTON_LABEL = "Copy the following snippet to cite these results"
@@ -81,6 +52,8 @@ CITATION_BUTTON_TEXT = r"""@article{jiang2024vlm2vec,
 SUBMIT_INTRODUCTION = """# Submit on MMEB Leaderboard Introduction
 
 ## ⚠ Please note that you need to submit the JSON file with the following format:
+
+### **TO SUBMIT V1 ONLY**
 ```json
 [
     {
@@ -88,15 +61,34 @@ SUBMIT_INTRODUCTION = """# Submit on MMEB Leaderboard Introduction
         <Optional>"URL": "<Model URL>",
         "Model Size(B)": 1000,
         "Data Source": Self-Reported,
-        "Overall": 50.0,
-        "Classification": 50.0,
-        "VQA": 50.0,
-        "Retrieval": 50.0,
-        "Grounding": 50.0
+        "V1-Overall": 50.0,
+        "I-CLS": 50.0,
+        "I-QA": 50.0,
+        "I-RET": 50.0,
+        "I-VG": 50.0
     }, 
 ]
 ```
-You may refer to the Github page for instructions about evaluating your model. 
+
+### **TO SUBMIT V2 ONLY**
+```json
+[
+    {
+        "Model": "<Model Name>",
+        <Optional>"URL": "<Model URL>",
+        "Model Size(B)": 1000,
+        "Data Source": Self-Reported,
+        "V2-Overall": 50.0,
+        "V-CLS": 50.0,
+        "V-QA": 50.0,
+        "V-RET": 50.0,
+        "V-VG": 50.0, 
+        "VisDoc": 50.0
+    }, 
+]
+```
+You are also welcome to submit both versions by including all the fields above! :) \n
+You may refer to the Github page for instructions about evaluating your model. \n
 Github link: https://github.com/TIGER-AI-Lab/VLM2Vec. \n
 Please send us an email at [email protected], attaching the JSON file. We will review your submission and update the leaderboard accordingly.
 """
@@ -113,24 +105,21 @@ def create_hyperlinked_names(df):
     df = df.apply(add_link_to_model_name, axis=1)
     return df
 
-def fetch_data(file: str) -> pd.DataFrame:
-    # fetch the leaderboard data from remote
-    if file is None:
-        raise ValueError("URL Not Provided")
-    url = f"https://huggingface.co/spaces/TIGER-Lab/MMEB/resolve/main/{file}"
-    print(f"Fetching data from {url}")
-    response = requests.get(url)
-    if response.status_code != 200:
-        raise requests.HTTPError(f"Failed to fetch data: HTTP status code {response.status_code}")
-    return pd.read_json(io.StringIO(response.text), orient='records', lines=True)
+# def fetch_data(file: str) -> pd.DataFrame:
+#     # fetch the leaderboard data from remote
+#     if file is None:
+#         raise ValueError("URL Not Provided")
+#     url = f"https://huggingface.co/spaces/TIGER-Lab/MMEB/resolve/main/{file}"
+#     print(f"Fetching data from {url}")
+#     response = requests.get(url)
+#     if response.status_code != 200:
+#         raise requests.HTTPError(f"Failed to fetch data: HTTP status code {response.status_code}")
+#     return pd.read_json(io.StringIO(response.text), orient='records', lines=True)
 
 def get_df(file="results.jsonl"):
-    df = fetch_data(file)
-    print(df.columns)
-    print('URL' in df.columns)
-    print(df)
+    df = pd.read_json(file, orient='records', lines=True)
     df['Model Size(B)'] = df['Model Size(B)'].apply(process_model_size)
-    df = df.sort_values(by=['Overall'], ascending=False)
+    df = df.sort_values(by=['V1-Overall'], ascending=False)
     df = create_hyperlinked_names(df)
     df['Rank'] = range(1, len(df) + 1)
     return df
@@ -185,7 +174,3 @@ def filter_columns_by_tasks(df, selected_tasks=None):
     
     available_columns = [col for col in selected_columns if col in df.columns]
     return df[available_columns]
-
-def get_task_choices():
-    return TASKS
-