run-0653
capability
citable URL: https://halobench.com/records/run-0653/ — this address never moves; the anchor /records/#run-0653 keeps resolving
date 2026-09-02 · window 2026-09-02T21:08:53Z → 2026-09-02T23:00:34Z
suite tau2-airline-full26 · config cfg-0182
depth 196,608 tokens · scoring deterministic
Metrics
| metric | value |
|---|---|
| tasks_passed | 21 |
| tasks_total | 25 |
Guard chain
waiver — Formal 4/4 needle guard at depth 8000 not run; pre-flight was a direct tool-call gate (valid get_weather tool call through /v1/chat/completions) plus the end-to-end serving validation of the same binary/model. Disclosed deviation.
Energy
metered window: eng-0277
Cited by — computed at build time, never stored
claims clm-0128
model pages qwen38-flash-next