This dataset contains approximately 500,000 chess positions with explanations for optimal moves, prepared for Supervised Fine-Tuning (SFT). The data is tokenized using Qwen/Qwen3-8B tokenizer.
Format
File: JSONL format (one JSON object per line)
Fields:
text: Full conversation in Qwen chat format with special tokens
fen: Position in FEN notation
move: Best move in UCI notation
explanation: Natural language explanation… See the full description on the dataset page: https://huggingface.co/datasets/aicrowd/ChessExplained.
