Add dataset card

README.md

@@ -0,0 +1,52 @@
+---
+license: other
+language:
+- en
+pretty_name: GCC 2025 chess policy datasets (searchless + gold)
+tags:
+- chess
+- distillation
+- policy
+- searchless
+- stockfish
+- uci
+- global-chess-challenge-2025
+---
+
+# GCC 2025 Chess Policy Datasets (searchless + gold)
+
+This repo contains the datasets used to train and evaluate a text-only chess policy model for the Global Chess Challenge 2025.
+
+## Files
+- `searchless.jsonl`: 2,048,000 searchless teacher-labeled positions.
+- `gold.jsonl`: 208,832 Stockfish-labeled positions (gold).
+- `labels_merged.jsonl`: merged labels (searchless + gold).
+- `policy_sft_v2.jsonl`: SFT training set with `gold_repeat=3` applied.
+- `labels_eval_5000.jsonl`: fixed 5k canary set for offline evaluation.
+- `notes/data_assets_20251231T220044Z.md`: dataset inventory and backup notes.
+
+## Schema
+### Common fields
+- `fen`: FEN string
+- `move_history_uci`: optional move history (UCI)
+- `legal_moves_uci`: list of legal moves (UCI)
+- `best_move`: labeled best move (UCI)
+- `label_source`: `searchless` or `stockfish`
+
+### Searchless-only fields
+- `teacher_model`, `teacher_checkpoint_step`
+- `top_moves`, `top_win_probs`
+- `bottom_moves`, `bottom_win_probs`
+- `best_win_prob`
+
+### Gold-only fields
+- `stockfish_depth`
+
+## Intended use
+- Supervised policy SFT for chess move selection.
+- Preference/ranking training using top/bottom move lists and win probabilities.
+
+## Notes
+- Prompt templates are stored in the training repo; this dataset only includes labels.
+- The canary set is fixed to enable paired comparisons across checkpoints.
+