---
base_model:
- Qwen/Qwen3.5-4B
language:
- en
license: apache-2.0
pipeline_tag: image-text-to-text
library_name: transformers
---

# V-Zero: Answer-Label-Free On-Policy Distillation with Contrastive Evidence Gating for Fine-Grained Visual Reasoning

This repository contains the V-Zero 4B checkpoint, introduced in the paper [V-Zero: Answer-Label-Free On-Policy Distillation with Contrastive Evidence Gating for Fine-Grained Visual Reasoning](https://arxiv.org/abs/2606.25319).

* **Code Repository:** [GitHub - eVI-group-SCU/V-Zero](https://github.com/eVI-group-SCU/V-Zero)

## Overview

V-Zero is an answer-label-free framework designed to improve fine-grained visual reasoning in multimodal large language models (MLLMs). It bypasses the need for costly external answer labels or manual verification rules by utilizing on-policy distillation combined with contrastive evidence gating. During training, the student model samples trajectories on the full image, while a teacher model replays those trajectories under paired positive (task-relevant) and negative (task-irrelevant) crops to evaluate student-sampled reasoning paths.

<p align="center">
  <img src="https://raw.githubusercontent.com/eVI-group-SCU/V-Zero/main/resource/method.png" alt="V-Zero Method Overview" width="100%">
</p>

## Citation

If you find this work useful for your research, please cite the paper:

```bibtex
@article{sun2026vzero,
  title={V-Zero: Answer-Label-Free On-Policy Distillation with Contrastive Evidence Gating for Fine-Grained Visual Reasoning},
  author={Sun, Haoxiang and Yi, Zhihang and Deng, Langxuan and Zhou, Yuhao and Jia, Peiqi and Zhao, Jian and Yuan, Li and Lv, Jiancheng and Wang, Tao},
  journal={arXiv preprint arXiv:2606.25319},
  year={2026}
}
```