ciregenzandClaude Opus 5 1db0a5d04e [eric] harness: vs_bench grades zero-card trials from the backend's own dryrun-report ...

A trial whose card was reaped before the readback could not be verified either way,
and excluding those rows was too charitable: it flipped openswarm 60.9% -> 100% and
McNemar p=0.02 -> 1.0 purely on my read timing. The backend's own dryrun-report
(composer/textboxes/filled) is the same bar the card readback applies, so it stands
in when the card is gone.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

2026-08-08 09:22:33 -07:00

..

anon_raw.jsonl

[eric] harness: OSW_CDP_PORT exposes Electron's CDP so a rival agent can share our browser

2026-08-07 21:14:58 -07:00

holdout_raw.jsonl

[eric] harness: OSW_CDP_PORT exposes Electron's CDP so a rival agent can share our browser

2026-08-07 21:14:58 -07:00

known_raw.jsonl

[eric] browser-v3: move the measurement harness into the repo, out of a temp dir

2026-08-06 13:27:26 -07:00

vs_raw.jsonl

[eric] harness: vs_bench grades zero-card trials from the backend's own dryrun-report

2026-08-08 09:22:33 -07:00