UniVG-R1-data

# UniVG-R1 Model Card
<a href='https://amap-ml.github.io/UniVG-R1-page/'><img src='https://img.shields.io/badge/Project-Page-Green'></a>
<a href='https://arxiv.org/abs/2505.14231'><img src='https://img.shields.io/badge/Paper-Arxiv-red'></a>
<a href='https://github.com/AMAP-ML/UniVG-R1'><img src='https://img.shields.io/badge/Code-GitHub-blue'></a>

## Model details
We propose UniVG-R1, a reasoning-guided MLLM for universal visual grounding, which leverages reinforcement learning to enhance reasoning across complex multi-image and multi-modal scenarios.

## Dataset details
We provide three JSON files as follows:
1. revised_MIG_bench.json: which contains our revised version of the MIG_bench.
2. stage1_cotsft.json: which contains the CoT-SFT data required for stage 1.
3. stage2_rl.json: which contains the RL data required for stage 2.
4. zero_shot_evaluation: which contains the bounding boxes annotations for zero-shot evaluation.

# UniVG-R1 模型卡片
<a href='https://amap-ml.github.io/UniVG-R1-page/'><img src='https://img.shields.io/badge/项目页面-Green'></a>
<a href='https://arxiv.org/abs/2505.14231'><img src='https://img.shields.io/badge/论文-Arxiv-red'></a>
<a href='https://github.com/AMAP-ML/UniVG-R1'><img src='https://img.shields.io/badge/代码-GitHub-blue'></a>

## 模型详情
我们提出UniVG-R1，一款面向通用视觉定位（universal visual grounding）的推理引导型多模态大语言模型（Multimodal Large Language Model, MLLM），该模型借助强化学习（reinforcement learning）以提升复杂多图像及多模态场景下的推理性能。

## 数据集详情
我们提供如下三个JSON文件：
1. revised_MIG_bench.json：包含我们修订后的MIG_bench数据集版本。
2. stage1_cotsft.json：包含第一阶段训练所需的思维链监督微调（Chain-of-Thought Supervised Fine-Tuning, CoT-SFT）数据。
3. stage2_rl.json：包含第二阶段训练所需的强化学习数据。
4. zero_shot_evaluation：包含零样本（Zero-shot）评估所需的边界框标注信息。