Board reconstruction as a state-tracking test: low loss doesn't prove real tracking
andreisavu · x · 2026-11-08
Parts 3-4 of the chess transformer thread cover the board objective and value/policy roles in search:
- Board objective: the setup predicts the contents of all 64 squares, but square occupancy omits castling rights, en passant and repetition history; the author cautions that low reconstruction loss alone doesn't establish reliable state tracking.
- Value in search: value describes the position from White's perspective; search uses W + 0.5D as White's expected score, evaluating a candidate by appending the move and scoring the resulting position — policy probability and expected score serve different roles.
More from Research
- The meta-question as agent swarms scale: can models beat gradient descent at its own game? — JensHonack · 2026-11-08
- Why can LLMs solve PhD-level math but still lose at chess? Reddit asks — MesloubFC · 2026-11-08
- WebHarbor launches: 153 mirrored real websites and 3,518 tasks for GUI web agents — HuaxiuYaoML · 2026-11-08
- METR Progress per R&D Dollar Fell 63% Even as Yearly Gains Doubled — sebkrier · 2026-11-08
- GenrobotAI's DAS Dex 2.0 captures 21-DoF hand motion plus 980 tactile points for robot learning — CyberRobooo · 2026-11-08
- fordycelab proposes benchmarking binder-design algorithms with Kd measurements on 24,000-VHH libraries — anshulkundaje · 2026-11-08