Datasetpublic

OpenMOSS-Team/GameQA-5K

In this repository, we specifically provide the 5k training samples from the complete GameQA-140K dataset used in our work for GRPO training of the models. Refer to our paper for details. And our code for training and evaluation is at https://github.com/tongjingqi/Code2Logic. 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… See the full description on the dataset page: https://huggingface.co/datasets/OpenMOSS-Team/GameQA-5K.

sourceHugging Facemitupdated 1y agoView on Hugging Face

2likes166downloads

Dataset Card

*In this repository, we specifically provide the 5k training samples from the complete [GameQA-140K](https://huggingface.co/datasets/Gabriel166/GameQA-140K) dataset used in our work for GRPO training of the models.*

Refer to our paper for details. And our code for training and evaluation is at https://github.com/tongjingqi/Code2Logic.

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] 🤗 [GameQA-140K Dataset] 🤗 [GameQA-InternVL3-8B ] 🤗 [GameQA-Qwen2.5-VL-7B] 🤗 [GameQA-LLaVA-OV-7B ]

Evaluation Results on General Vision BenchMarks

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

It's also found that getting trained on the 5k samples from our GameQA dataset can lead to better results than on multimodal-open-r1-8k-verified.

News

—We've open-sourced the *three* models trained with GRPO on GameQA on Huggingface.