Data S1: game records, engine evaluations and codings for "Checking language-model reasoning against ground truth in chess played through a robot arm"
收藏资源简介:
Supplementary Data S1 (related to Figures 2-5, Figures S2 and S3, and Tables S1 and S2). The records of the 21 games counted in the paper and of the void full-size game F: 8 games of Gardner minichess moved by a UFACTORY Lite 6 arm (4-8 October 2026), 4 minichess games played in software (13-14 October 2026) and 10 full-size games on a stone set moved by the same arm (10-23 October 2026). For every model call: the prompt, the reply, the agent's full output with token counts, and reasoning text where the server returned it; for every game: launch records (commands, versions, script hashes, model identity), incidents, board checks and arm logs. Also the engine reference evaluations, the reasoning and physical-deviation codings and the values plotted in the figures. Camera frames, video, depth images and point clouds are not included. README.md describes every folder; MANIFEST.tsv lists every file with its size and SHA-256.



