README.md

---
license: apache-2.0
datasets:
- Code2Logic/GameQA-140K
- Code2Logic/GameQA-5K
base_model:
- llava-hf/llava-onevision-qwen2-7b-ov-hf
---

***This model (GameQA-LLaVA-OV-7B) results from training LLaVA-OV-7B with GRPO solely on our [GameQA-5K](https://huggingface.co/datasets/Code2Logic/GameQA-5K) (sampled from the full [GameQA-140K](https://huggingface.co/datasets/Gabriel166/GameQA-140K) dataset).***

# Evaluation Results on General Vision BenchMarks

<div align=center><img src="https://raw.githubusercontent.com/tongjingqi/Code2Logic/refs/heads/main/assets/evaluation_results_on_general_vision_benchmarks.png"></div>

***(The inference and evaluation configurations were unified across both the original open-source models and our trained models.)***

# Code2Logic: Game-Code-Driven Data Synthesis for Enhancing VLMs General Reasoning

This is the first work, to the best of our knowledge, that leverages ***game code*** to synthesize multimodal reasoning data for ***training*** VLMs. Furthermore, when trained with a GRPO strategy solely on **GameQA** (synthesized via our proposed **Code2Logic** approach), multiple cutting-edge open-source models exhibit significantly enhanced out-of-domain generalization.

[[📖 Paper](https://arxiv.org/abs/2505.13886)] [[🤗 GameQA-140K Dataset](https://huggingface.co/datasets/Gabriel166/GameQA-140K)] [[🤗 GameQA-5K Dataset](https://huggingface.co/datasets/Code2Logic/GameQA-5K)] [[🤗 GameQA-InternVL3-8B](https://huggingface.co/Code2Logic/GameQA-InternVL3-8B) ] [[🤗 GameQA-Qwen2.5-VL-7B](https://huggingface.co/Code2Logic/GameQA-Qwen2.5-VL-7B)] [[🤗 GameQA-LLaVA-OV-7B](https://huggingface.co/Code2Logic/GameQA-llava-onevision-qwen2-7b-ov-hf) ]

## News

* We've open-sourced the ***three*** models trained with GRPO on GameQA on [Huggingface](https://huggingface.co/Code2Logic).