eng-0175
window reconstructed
citable URL: https://halobench.com/records/eng-0175/ — this address never moves; the anchor /records/#eng-0175 keeps resolving
802.7 Whmean 153.8 W − 10.1 W idle floor (box-idle-all-empty) = 143.7 W attributable
window 2026-08-10T03:06:47Z → 2026-08-10T08:19:49Z · provenance reconstructed
method wall-meter · aihydra energy sensor via Home Assistant (counter-difference on sensor.hardware_ai_hydra_energy, raw ~10s-resolution history samples, linearly interpolated to the transcript-derived window) · no contention
per task 100.3336 Wh · 3.0401 p at 30.3 p/kWh
covers run-0101
aihydra qwen35-122b tau2-bench-airline 9-scored-task arm window (2026-08-10T03:06:47Z..2026-08-10T08:19:49Z, 18782s), taken from run-0101's started_at/ended_at. run-0101 predates matching run-meta instrumentation for this arm, so the window is reconstructed from transcript timestamps rather than a recorded meter window. Counter interpolated from raw ~10s-resolution HA history samples pulled fresh 2026-08-18 as part of the run-0101/run-0102/run-0111 batch: start 5.740920207 kWh, end 6.543588647 kWh. The transcript contains a long internal idle gap between task 6 and the later task 7/8/9 block; the gap is included end-to-end as wall-clock session energy. wh_per_task is Wh-per-CORRECT-ANSWER: wh_total / tasks_passed (8 of 9 scored tasks, run-0101.metrics) = 100.3336 Wh, 3.0401 pence/correct at the standing 30.3 p/kWh tariff. This arm ran an unpinned self-play user simulator (user_llm is the same qwen35-122b path as the agent), so it remains provisional per clm-0043.
Wh, never joules — read as a cumulative-counter difference across the window; delta over the idle baseline is the only figure that means anything
Cited by — computed at build time, never stored
claims clm-0042