update

app.py

@@ -1,11 +1,9 @@
 import gradio as gr
 from huggingface_hub import InferenceClient
-import time
 
-# Initialize the client with your model
+# Initialize the client
 client = InferenceClient("zhangchenxu/TinyV-1.5B")
 
-# The prompt template for the LLM verifier
 LV_PROMPT = """
 You are an AI tasked with identifying false negatives in answer verification. A false negative occurs when a model's answer is essentially correct but is marked as incorrect due to minor discrepancies or formatting issues. Your job is to analyze the given question, ground truth answer, and model answer to determine if the model's answer is actually correct despite appearing different from the ground truth.
 
@@ -67,25 +65,15 @@ EXAMPLES = [
     }
 ]
 
-# Main verification function
 def verify_answer(question, ground_truth, model_answer, temperature, top_p, max_tokens):
-    # Format the prompt with user inputs
     prompt = LV_PROMPT.format(
         question=question,
         ground_truth=ground_truth,
         model_answer=model_answer
     )
-    
-    # Prepare the message format required by the API
-    messages = [
-        {"role": "user", "content": prompt}
-    ]
-    
-    # Initialize response
+    messages = [{"role": "user", "content": prompt}]
     response_text = ""
-    
     try:
-        # Stream the response for better UX
         for message in client.chat_completion(
             messages,
             max_tokens=max_tokens,
@@ -100,7 +88,6 @@ def verify_answer(question, ground_truth, model_answer, temperature, top_p, max_
     except Exception as e:
         yield f"Error: {str(e)}"
 
-# Function to load an example when its button is clicked
 def load_example(example_index):
     example = EXAMPLES[example_index]
     return (
@@ -112,108 +99,42 @@ def load_example(example_index):
         example["tokens"]
     )
 
-# Create the Gradio interface
-with gr.Blocks(theme=gr.themes.Soft(primary_hue="blue", font=[gr.themes.GoogleFont("Inter"), "ui-sans-serif", "system-ui", "sans-serif"]), title="TinyV") as demo:
-    # Header with title and description
-    with gr.Row():
-        with gr.Column():
-            gr.Markdown(
-                """
-                # TinyV - Answer Verification Tool
-                
-                This tool verifies if an answer is correct compared to a ground truth answer for RL.
-                """
-            )
+with gr.Blocks(theme=gr.themes.Soft(primary_hue="blue")) as demo:
+    gr.Markdown("## 🧠 TinyV - Answer Verification Tool\nThis tool verifies model-generated answers for correctness.")
     
-    # Main interface
     with gr.Row():
         with gr.Column(scale=1):
-            gr.Markdown(
-                """
-                ## How to Use
-                
-                1. Enter the question in the first box
-                2. Enter the ground truth answer
-                3. Enter the model's answer to verify
-                4. Adjust model parameters if needed
-                5. Click "Verify Answer" to see the result
-                
-                ### What this tool does
-                
-                This tool determines if a model's answer is semantically correct compared to a ground truth answer using a fine-tuned LLM.
-                
-                The model analyzes both answers and returns:
-                - **True** if the model answer is correct
-                - **False** if the model answer is incorrect
-                
-                ### API Usage Example
-                ```python
-                from gradio_client import Client
-                
-                client = Client("zhangchenxu/TinyV")
-                result = client.predict(
-                    question="Determine all real values of $x$ for which $(x+8)^{4}=(2 x+16)^{2}$.",
-                    ground_truth="-6,-8,-10",
-                    model_answer="-10, -8, -6",
-                    temperature=0.3,
-                    top_p=0.95,
-                    max_tokens=1,
-                    api_name="/verify_answer"
-                )
-                print(result)
-                ```
-                """
-            )
-            
-            # Model parameters (hidden in a collapsible section)
-            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=1, maximum=256, value=1, step=1, label="Max Tokens")
-                
-        with gr.Column(scale=1):
-            gr.Markdown("## Input")
-            question = gr.Textbox(lines=3, label="Question", placeholder="Enter the question here...")
-            ground_truth = gr.Textbox(lines=5, label="Ground Truth Answer", placeholder="Enter the correct answer here...")
-            model_answer = gr.Textbox(lines=5, label="Model Answer", placeholder="Enter the answer to verify here...")
-            
-            # Examples section as buttons
-            gr.Markdown("### Try an example:")
+            question = gr.Textbox(lines=3, label="📘 Question")
+            ground_truth = gr.Textbox(lines=3, label="✅ Ground Truth Answer")
+            model_answer = gr.Textbox(lines=3, label="🤖 Model Answer")
+
+            gr.Markdown("### 🔍 Try Examples:")
+            example_buttons = []
             with gr.Row():
-                example_buttons = []
-                for i, example in enumerate(EXAMPLES):
-                    btn = gr.Button(example["name"], size="sm")
-                    example_buttons.append(btn)
-                    # Connect each button to the load_example function
+                for i, ex in enumerate(EXAMPLES):
+                    btn = gr.Button(ex["name"], size="sm")
                     btn.click(
                         fn=lambda idx=i: load_example(idx),
                         outputs=[question, ground_truth, model_answer, temperature, top_p, max_tokens]
                     )
-            
-            verify_btn = gr.Button("Verify Answer", variant="primary")
-            
-            gr.Markdown("## Result")
-            result = gr.Textbox(label="Verification Result", placeholder="Result will appear here...", lines=5)
-                
-    # Connect the interface to the verification function
+                    example_buttons.append(btn)
+
+        with gr.Column(scale=1):
+            with gr.Accordion("⚙️ Advanced Settings", open=False):
+                temperature = gr.Slider(0, 1, value=0.3, step=0.1, label="Temperature")
+                top_p = gr.Slider(0.1, 1.0, value=0.95, step=0.05, label="Top-p")
+                max_tokens = gr.Slider(1, 128, value=2, step=1, label="Max Tokens")
+
+            verify_btn = gr.Button("✅ Verify Answer", variant="primary")
+            result = gr.Textbox(label="🧾 Verification Result", lines=5, placeholder="Result will appear here...")
+
     verify_btn.click(
-        verify_answer,
+        fn=verify_answer,
         inputs=[question, ground_truth, model_answer, temperature, top_p, max_tokens],
         outputs=result
     )
-    
-    # Run verification when an example is loaded (optional)
-    for btn in example_buttons:
-        btn.click(
-            fn=verify_answer,
-            inputs=[question, ground_truth, model_answer, temperature, top_p, max_tokens],
-            outputs=result,
-            _js="() => {setTimeout(() => document.querySelector('#verify-btn').click(), 100)}",
-            queue=False
-        )
 
-# Define the public API
-demo.queue()
 # Launch the app
+demo.queue()
 if __name__ == "__main__":
     demo.launch()