dwidlee/systemone-lite-phase2
systemone-lite-phase2 Typed System One distill rows (task / state / instructions / criteria / label_alias) for systemone-lite. Revision (2026-09-26) Paired with Hub model revision action-v2-qwen (dwidlee/systemone-lite-0.5b). Change Detail Chess staged_v1 — piece + destination, option caps ≤8 Spatial action_v2 — legal-only actions; Connect4 drop≤3 + win_now Sokoban test Deadlock alerts balanced 75/75 yes/no; remap_alert_prob≈0.35 Postmortem:… See the full description on the dataset page: https://huggingface.co/datasets/dwidlee/systemone-lite-phase2.
Dataset card: action-v2 / staged chess revision
action-v2 + staged chess + balanced sokoban eval (0% train∩test)
Dataset card: document zero train/test leakage
Zero-leakage rebuild: train/test disjoint (vocab/topic/state) + nlp_cloze
Upload dataset
Upload dataset
Upload dataset
Upload dataset
Upload dataset
Upload dataset
initial commit