A natural-language AlphaZero: teaching LMs to play and explain chess
ChengleiSi · x · 2026-10-05
An NLP researcher calls this the most profound paper of his PhD: getting language models to play chess and explain their moves.
- The method combines architectural and algorithmic innovations—a natural-language analogue of the AlphaZero algorithm.
- Beyond playing, the model can explain its reasoning.
- The authors claim the approach generalizes to many domains beyond chess.
A substantive research release worth watching for self-play and language-space reasoning work.
More from Research
- Optimized open-source Argus hits SOTA on WGO-Bench, lifting semantic F1 from 29.8% to 50.2% — _sonith · 2026-10-06
- Alison Gopnik's 'Explanation as Orgasm' Hypothesis Goes Viral — stevenstrogatz · 2026-10-06
- First NeurIPS 2026 Agent Behavior Workshop Accepts 149 Papers — DanielKhashabi · 2026-10-06
- Black Forest Labs open-sources 7B FLUX 3 Action, tops RoboLab-120 at 3.95x speed — dl_weekly · 2026-10-06
- SwiLA explained: a mixture of J linear regressions that reduces to DeltaNet at J=1 — YouJiacheng · 2026-10-06
- MIT paper: stating rules beats reasoning prompts for LLM agents — Gemma's underbid gap shrinks from $5.30 to $0.30 — rohanpaul_ai · 2026-10-06