Spaces:
Running
Running
Commit History
Today
-
Gemini CLI: thought summaries filed as messages by Harbor's old converter, counted as reasoning (summary only)
Verified signature05c21f3 -
Trace completeness: share of output tokens in the trace; why not (withheld by provider, not kept by harness, lost in conversion, summary only); counted/inferred basis; undercounted
Verified signaturefe0df2c -
Output tokens as shares: tool calls, messages, reasoning shown, reasoning hidden
Verified signature00853bc
Yesterday
-
Cost estimate: only a fit that reproduces recorded costs prices trials; unpriced wording; PR-aware row page
Verified signature8cecfac -
Our estimate (leads + fallback discounted), output/hidden reasoning bar, cost estimate, regrouped gaps, waffle flags, Harbor Hub trial links
Verified signatureb8a3c5c -
Viewer: fold many argument-value fields per call (atif-scan 43ae555)
Verified signature77fbc9e -
Quieter design: leaderboard + row pages + compare; full trajectories for highlighted trials (atif-scan 90c8670 --viewer-split)
Verified signature5d9b93c -
Models served/fallback, reasoning and output-token accounting, drill-down to trials, tooltip fix
Verified signature8c073f5 -
Data holes & accounting; leaderboard rows scored as listed (atif-scan v0.8.17-2-g4fbd8b8)
Verified signature2cbfd05 -
TB2.1 annotated: 22 leaderboard rows, atif-scan v0.8.17 highlights and trajectories
Verified signatureedc735f