Update README.md

README.md

@@ -1,37 +1,10 @@
 ---
 license: mit
 ---
-## 💡 Overview
+# V-Thinker: Interactive Thinking with Images
 
-> *"The soul never thinks without an image." — Aristotle*
+The model checkpoint of ARPO is released for the paper V-Thinker: Interactive Thinking with Images
 
-**V-Thinker** is a general-purpose multimodal reasoning assistant that enables **Interactive Thinking with Images** through end-to-end reinforcement learning. Unlike traditional vision-language models, V-Thinker actively **interacts** with visual content—editing, annotating, and transforming images to simplify complex problems.
-```bash
-import torch
-import os
-import json
-import argparse
-from transformers import Qwen2_5_VLForConditionalGeneration, AutoProcessor, AutoConfig, Qwen3VLForConditionalGeneration
-from tqdm import tqdm
-from utils import run_evaluation # Assuming you have this utility function
-MODEL_PATH=""
+## Abstract
 
-config = AutoConfig.from_pretrained(MODEL_PATH)
-model = Qwen2_5_VLForConditionalGeneration.from_pretrained(
-    MODEL_PATH,
-    device_map="auto", # "auto" works perfectly with CUDA_VISIBLE_DEVICES
-    config=config
-)
-processor = AutoProcessor.from_pretrained(MODEL_PATH)
-
-question_text = "Question: Hint: Please answer the question and provide the final answer at the end.\nQuestion: How many lines of symmetry does this figure have?\n\n\nPlease provide the final answer in the format <answer>X</answer>"
-image_path = "./224.png"
-    
-    # Construct the full, normalized image pat
-final_assistant_response, final_answer, aux_path = run_evaluation(question_text, image_path, "./", model, processor)
-print("Model Response")
-print(final_answer)
-print("auxiliary path")
-print(final_answer)   
-
-```
+Empowering Large Multimodal Models (LMMs) to deeply integrate image interaction with long-horizon reasoning capabilities remains a long-standing challenge in this field. Recent advances in vision-centric reasoning explore a promising “Thinking with Images” paradigm for LMMs, profoundly shifting from image-assisted reasoning to image-interactive thinking. While this milestone enables models to focus on fine-grained image regions, progress remains constrained by narrow visual tool spaces and task-specific workflow designs. To bridge this gap, we present V-Thinker, a general-purpose multimodal reasoning assistant that enables interactive, vision-centric thinking through end-to-end reinforcement learning. V-Thinker comprises two key components: (1) a Data Evolution Flywheel that automatically synthesizes, evolves, and verifies interactive reasoning datasets across three dimensions — diversity, quality, and difficulty; and (2) a Visual Progressive Training Curriculum that first aligns perception via point-level supervision, then integrates interactive reasoning through a two-stage reinforcement learning framework. Furthermore, we introduce VTBench, an expert-verified benchmark targeting vision-centric interactive reasoning tasks. Extensive experiments demonstrate that V-Thinker consistently outperforms strong LMM-based baselines in both general and interactive reasoning scenarios, providing valuable insights for advancing image-interactive reasoning applications.