Upload 4 files

LICENSE.txt

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2025 Zhihui Chen
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

README.md

@@ -1,14 +1,30 @@
----
-title: DivScore
-emoji: 🌍
-colorFrom: indigo
-colorTo: red
-sdk: gradio
-sdk_version: 5.33.1
-app_file: app.py
-pinned: false
-license: mit
-short_description: 'DivScore: Zero-Shot Detection of LLM-Generated Text in Speci'
----
-
-Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
+---
+title: DivScore AI Text Detector
+emoji: 🔍
+colorFrom: blue
+colorTo: purple
+sdk: gradio
+sdk_version: 4.19.2
+app_file: app.py
+pinned: false
+---
+
+# DivScore AI Text Detector
+
+This Space demonstrates the DivScore model for detecting AI-generated text. The model uses a novel zero-shot detection framework that leverages normalized entropy-based scoring to identify AI-generated content.
+
+## Features
+- Real-time AI text detection
+- Detailed analysis scores (DivScore, Entropy, CE)
+- Confidence level indication
+- Easy-to-use interface
+
+## How to Use
+1. Enter or paste your text in the input box
+2. Click "Submit" to analyze
+3. View the results showing whether the text is likely AI-generated
+
+## Technical Details
+- Based on Mistral-7B model
+- Uses DivScore detection framework
+- Threshold: 0.15 (scores below this are likely AI-generated)

app.py

@@ -0,0 +1,94 @@
+import gradio as gr
+from divscore import DivScore
+import torch
+import os
+
+# Set environment variables for Hugging Face
+os.environ["TOKENIZERS_PARALLELISM"] = "false"
+
+# Initialize the DivScore detector with loading state
+def load_model():
+    try:
+        detector = DivScore(
+            generalLM_name_or_path="mistralai/Mistral-7B-v0.2",
+            enhancedLM_name_or_path="RichardChenZH/DivScore_combined",
+            device="cuda:0" if torch.cuda.is_available() else "cpu",
+            use_bfloat16=True  # Use bfloat16 for better memory efficiency
+        )
+        return detector
+    except Exception as e:
+        print(f"Error loading model: {str(e)}")
+        return None
+
+# Global variable for the detector
+detector = None
+
+def detect_ai_text(text):
+    """
+    Detect if the input text is AI-generated using DivScore.
+    Returns a tuple of (score, is_ai_generated, confidence)
+    """
+    global detector
+    
+    # Initialize detector if not already done
+    if detector is None:
+        detector = load_model()
+        if detector is None:
+            return "Error: Failed to load the model. Please try again later.", False, 0.0
+    
+    if not text.strip():
+        return "Please enter some text to analyze.", False, 0.0
+    
+    try:
+        score, entropy_score, ce_score = detector.compute_score(text)
+        
+        # Based on the paper's findings, we use 0.15 as the threshold
+        is_ai_generated = score < 0.15
+
+        result = f"DivScore: {score:.4f}\nEntropy Score: {entropy_score:.4f}\nCE Score: {ce_score:.4f}"
+        return result, is_ai_generated
+        
+    except Exception as e:
+        return f"Error occurred: {str(e)}", False, 0.0
+
+# Create the Gradio interface with loading state
+with gr.Blocks(title="DivScore AI Text Detector") as demo:
+    gr.Markdown("""
+    # DivScore AI Text Detector
+    
+    This demo uses the DivScore model to detect if text was generated by an AI model.
+    Enter your text below to analyze it.
+    
+    **Note:** The model may take a few moments to load on first use.
+    """)
+    
+    with gr.Row():
+        with gr.Column():
+            text_input = gr.Textbox(
+                label="Input Text",
+                placeholder="Enter text to analyze...",
+                lines=5
+            )
+            submit_btn = gr.Button("Analyze Text")
+        
+        with gr.Column():
+            result_output = gr.Textbox(label="Analysis Results")
+            ai_generated = gr.Checkbox(label="AI Generated", interactive=False)
+    
+    gr.Examples(
+        examples=[
+            ["The quick brown fox jumps over the lazy dog."],
+            ["Based on the analysis of the data, we can conclude that the implementation of the new protocol has resulted in a statistically significant improvement in patient outcomes."]
+        ],
+        inputs=text_input
+    )
+    
+    submit_btn.click(
+        fn=detect_ai_text,
+        inputs=text_input,
+        outputs=[result_output, ai_generated]
+    )
+
+if __name__ == "__main__":
+    demo.queue()  # Enable queuing for better handling of multiple requests
+    demo.launch() 

requirements.txt

@@ -0,0 +1,5 @@
+gradio>=4.0.0
+torch>=2.0.0
+transformers>=4.30.0
+peft>=0.4.0
+accelerate>=0.20.0