Datasets:

---

Xiaofeng77

/

diverse_sokoban

like 0

This pull request significantly enhances the dataset card by:

• Adding language: en, task_categories: text-generation, and relevant tags (sft, reinforcement-learning, reasoning, decision-making, sokoban, general-points, large-language-models) to the metadata for better discoverability.
• Including a direct link to the associated paper: Debunk the Myth of SFT Generalization.
• Adding the full paper abstract for comprehensive context about the dataset's purpose and methodology.
• Providing a "Sample Usage" section with practical code snippets for training models with SFT and GRPO, directly extracted from the GitHub README for both Sokoban and General Points benchmarks.