Published finding
hy4-preview@deepinfra: thinking-token use rose on the instruction following probe set
Movement on the daily battery. This model measured against its own witnessed baseline.
967 tok→3,028 tok
| model | hy4-preview@deepinfra |
|---|---|
| identity | alias · a floating name the provider may re-point; a change here means what is served under this name changed |
| dimension | Instruction following |
| metric | thinking_tokens_mean |
| decision rule | Relative shift past the serving-tripwire threshold |
| first flagged | 2026-10-21 · latest reading |
| coverage | 6 call errors on this reading |
Pattern: thinking-token use and the pass rate on the same probes moved together (1.00 before, 0.87 at onset). Accuracy shifting alongside token use is more consistent with a different model behind the name than with an effort or routing change alone. Stated as a pattern, not a cause.
Trace the evidence
The witnessed reading for this date and battery, and the Rekor witness bundle. The full series is in series.json.
This entry records an observation under the stated decision rule. Watches remain unconfirmed; latency is operational context. The instrument does not establish the cause of a change.