Text-only reasoning pairs and logic-based question-answer sets synthesized from game code across multiple game environments. This data utilizes game mechanics to facilitate training and evaluation of general reasoning in models via the Code2Logic framework.
Use Cases
- Fine-tune models on complex logic tasks using the synthesized game-based question and answer text
- Evaluate the reasoning performance of LLMs against game-logic scenarios derived from actual code
- Train models using GRPO strategies to optimize reasoning trajectories based on the provided text data
Strengths
- Synthesized using the Code2Logic framework which converts game mechanics into logical reasoning data
- Provides a pure-text version of the multimodal GameQA dataset for text-only model training
- Designed for use with the GRPO (Group Relative Policy Optimization) strategy to enhance model logic