datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
move-intent-final-evidence
Move Intent final evidence
This repository preserves the pinned evaluator, raw candidates, raw judge replies, run specifications, receipts, and reports behind the Move Intent submission.
eval.py is the public grading entry point. Its immutable repository revision is recorded in the final submission manifest.
reports/prompt-ceiling.md is the official API-judge ablation and uses the same pinned rubric as the final comparison and curve. reports/prompt-ceiling-subscription-audit.md… See the full description on the dataset page: https://huggingface.co/datasets/rrodolfo0/move-intent-final-evidence.move-intent-v2-final
Move Intent v2 final
This is the exact submitted transcript-only Move Intent corpus. Given one finalized English chess transcript and no board state, the target is one canonical move-interpretation/v2 compact value or UNKNOWN.
The target is Move Interpretation, not SAN. The model never receives FEN, legal moves, move history, resolver candidates, or SAN. A host-owned Move Resolver checks the interpretation against chess.js legal moves; chess.js derives legality, playable… See the full description on the dataset page: https://huggingface.co/datasets/rrodolfo0/move-intent-v2-final.