nuriyev/

chess-reasoning

A small supervised fine-tuning dataset for chess move reasoning.

This dataset is designed for preliminary SFT to align language models with a structured reasoning format before further refinement via prolonged GRPO with Stockfish rewards.

Each example contains:

fen: Current board position in FEN notation reasoning: Brief chain-of-thought: "I see" X threat/opportunity, "so I will do" Y uci_move: The next predicted best… See the full description on the dataset page: https://huggingface.co/datasets/nuriyev/chess-reasoning.