BinProp scaffold nearly triples LLMs' chess-explanation scores by writing analysis code
OfirPress · x · 2026-10-14
Ori Press's team released BinProp, a scaffold that lets language models explain why a chess move is good by writing code that analyzes the board, grounded in 200k real-game board states with engine analysis.
On the team's chess-explanation benchmark, BinProp nearly triples LLMs' scores, and it generalizes to other games. A website lets users try it on their own games.
More from Research
- For small models, ipython is all you need: single-tool harness ties production-grade scaffolds on TB-Lite — yb2698 · 2026-10-14
- 20-trial repeat experiments: OpenHands stays most stable across sampling settings — yb2698 · 2026-10-14
- 72 trials per model-harness pair on TB-Lite: no clear winner emerges — yb2698 · 2026-10-14
- 2015 panel resurfaces: Tao vs Kontsevich on AI replacing mathematicians — Singularitarian · 2026-10-14
- One Night of Sleep Data Plus AI Predicts Future Disease and Mortality — EricTopol · 2026-10-14
- AI biomolecular design startup Boltz lands on Disruptors 60 list — GabriCorso · 2026-10-14