khoilamalphaai/chess-coach-move-review
Chess coach move-review SFT dataset Supervised fine-tuning data for one specific, trained behavior: given a chess position and the student's rating tier (Beginner, Intermediate, or Advanced), select the tier-appropriate instructive move and tag it with a short principle, for example "Nf3, develop toward the center." That single move choice is the trained objective, and it is deterministically checkable. The plain-English explanation rendered beside the move is a secondary… See the full description on the dataset page: https://huggingface.co/datasets/khoilamalphaai/chess-coach-move-review.
Clarify dataset card: behavior, build, schema, lineage, caveats
docs: standardize unbiased head-to-head on the reproducible 56-24-12 over 92 diverging (was the non-reproducing 56-28-12/96); asserted by reproduce_v4
docs: label 51-5-6 as selection-conditioned subset + add unbiased 56-28-12
docs(card): correct v4 eval line to strict numbers (tier-fit 0.767 vs base 0.347)
Register v4 config as default; document v4/train.jsonl as the shipped SFT set (keep v1-v3)
Revamp model/dataset card to match research standards
Upload 4b-iter1/manifest.json with huggingface_hub
Upload 4b-iter1/valid_4b_iter1.jsonl with huggingface_hub
Upload 4b-iter1/train_4b_iter1.jsonl with huggingface_hub
backup valid_v4.jsonl
backup train_v4.jsonl
Add v3 config + v2->v3 section to dataset card
Add v3 validation split
Add v3 train split
Reference the definitive 803-position, 14-model eval (model trained on this v2 data leads the moat) + link gap803
Namespace into v1/ + v2/; add v2 training data (2,457/129); update card
Upload README.md with huggingface_hub
Upload validation.jsonl with huggingface_hub
Upload train.jsonl with huggingface_hub
initial commit