clm-0078
measured-herehigh ●●●
citable URL: https://halobench.com/records/clm-0078/ — this address never moves; the anchor /records/#clm-0078 keeps resolving
Deep-Thought-Posttrain's full 26-task tau2 airline run (run-0372) recorded tool_call_messages = 0 across all 26 simulations — the same SMOKE validity gate (protocol.json comparison_rules, added 2026-08-14) that flagged npu-lfm2's vacuous 1.00 mean reward as invalid on this site's first NPU screen.
verified 2026-08-17 · volatility low
evidence run-0372
Note — the record's own working
Digit-backing claim for the model page's compare caveat (site-design-v2 §7: no number appears in rendered prose without a cited claim containing it verbatim).
Cited by — computed at build time, never stored
model pages deep-thought-posttrain