update

app.py

@@ -29,7 +29,6 @@ def verify_answer(question, ground_truth, model_answer, temperature, top_p, max_
     
     # Prepare the message format required by the API
     messages = [
-        {"role": "system", "content": "You are a helpful AI assistant that verifies answers."},
         {"role": "user", "content": prompt}
     ]
     
@@ -53,12 +52,10 @@ def verify_answer(question, ground_truth, model_answer, temperature, top_p, max_
         yield f"Error: {str(e)}"
 
 # Create the Gradio interface
-with gr.Blocks(theme=gr.themes.Soft(primary_hue="blue"), title="Answer Verification Tool") as demo:
+with gr.Blocks(theme=gr.themes.Soft(primary_hue="blue", font=[gr.themes.GoogleFont("Inter"), "ui-sans-serif", "system-ui", "sans-serif"]), title="Answer Verification Tool") as demo:
     # Header with title and description
     with gr.Row():
-        with gr.Column(scale=1):
-            gr.Image("https://huggingface.co/front/assets/huggingface_logo.svg", scale=1, show_label=False, height=64)
-        with gr.Column(scale=5):
+        with gr.Column():
             gr.Markdown(
                 """
                 # Answer Verification Tool
@@ -99,8 +96,8 @@ with gr.Blocks(theme=gr.themes.Soft(primary_hue="blue"), title="Answer Verificat
                     model_answer="Paris is the capital of France.",
                     temperature=0.3,
                     top_p=0.95,
-                    max_tokens=128,
-                    api_name="/verify"
+                    max_tokens=1,
+                    api_name="/verify_answer"
                 )
                 print(result)
                 ```
@@ -111,7 +108,7 @@ with gr.Blocks(theme=gr.themes.Soft(primary_hue="blue"), title="Answer Verificat
             with gr.Accordion("Advanced Settings", open=False):
                 temperature = gr.Slider(minimum=0.0, maximum=1.0, value=0.3, step=0.1, label="Temperature")
                 top_p = gr.Slider(minimum=0.1, maximum=1.0, value=0.95, step=0.05, label="Top-p (nucleus sampling)")
-                max_tokens = gr.Slider(minimum=32, maximum=512, value=128, step=32, label="Max Tokens")
+                max_tokens = gr.Slider(minimum=32, maximum=512, value=1, step=32, label="Max Tokens")
                 
         with gr.Column(scale=1):
             gr.Markdown("## Input")
@@ -131,56 +128,67 @@ with gr.Blocks(theme=gr.themes.Soft(primary_hue="blue"), title="Answer Verificat
         outputs=result
     )
     
-    # Examples
+    # Examples section - improved styling
     with gr.Accordion("Examples", open=True):
-        gr.Examples(
-            examples=[
-                [
-                    "What is the capital of France?",
-                    "The capital of France is Paris.",
-                    "Paris is the capital of France.",
-                    0.3,
-                    0.95,
-                    128,
-                ],
-                [
-                    "What is 2+2?",
-                    "4",
-                    "The answer is 4.",
-                    0.3,
-                    0.95,
-                    128,
+        with gr.Row():
+            gr.Examples(
+                examples=[
+                    [
+                        "What is the capital of France?",
+                        "The capital of France is Paris.",
+                        "Paris is the capital of France.",
+                        0.3,
+                        0.95,
+                        2,
+                    ],
+                    [
+                        "What is 2+2?",
+                        "4",
+                        "Four.",
+                        0.3,
+                        0.95,
+                        2,
+                    ],
                 ],
-                [
-                    "When was the Declaration of Independence signed?",
-                    "July 4, 1776",
-                    "The Declaration of Independence was signed on July 4th, 1776.",
-                    0.3,
-                    0.95,
-                    128,
-                ],
-                [
-                    "List the first three planets from the sun.",
-                    "Mercury, Venus, Earth",
-                    "The first three planets from the sun are Mercury, Venus, and Earth.",
-                    0.3,
-                    0.95,
-                    128,
+                inputs=[question, ground_truth, model_answer, temperature, top_p, max_tokens],
+                outputs=result,
+                fn=verify_answer,
+                cache_examples=True,
+            )
+        
+        with gr.Row():
+            gr.Examples(
+                examples=[
+                    [
+                        "When was the Declaration of Independence signed?",
+                        "July 4, 1776",
+                        "The Declaration of Independence was signed on July 4th, 1776.",
+                        0.3,
+                        0.95,
+                        2,
+                    ],
+                    [
+                        "List the first three planets from the sun.",
+                        "Mercury, Venus, Earth",
+                        "The first three planets from the sun are Mercury, Venus, and Earth.",
+                        0.3,
+                        0.95,
+                        2,
+                    ],
                 ],
-            ],
-            inputs=[question, ground_truth, model_answer, temperature, top_p, max_tokens],
-            outputs=result,
-        )
+                inputs=[question, ground_truth, model_answer, temperature, top_p, max_tokens],
+                outputs=result,
+                fn=verify_answer,
+                cache_examples=True,
+            )
     
     # Add footer with extra information
-    with gr.Row():
+    with gr.Row(equal_height=True):
         gr.Markdown(
             """
             ### About
-            This tool uses the zhangchenxu/TinyV-1.5B model to verify answers.
-            
-            The verification is based on semantic similarity rather than exact matching,
-            allowing for different phrasings and formats of the same correct answer.
+            This tool uses the zhangchenxu/TinyV-1.5B model to verify the correctness of the answers,
+            allowing for different phrasings of correct answers.
             """
         )