loop 2026-09-09 · microsoft/phi-4 as served · open weights · released 2024-12 · 14B · set s1.4-live · cases v1.7 · conditions v1.0 · ground truth as of 2026-08-09
From no context to real context
5 repeats per case
whiskers are 95% intervals · each step is measured from the stop before it, not from zero
Try the other contexts
stored · no model call
1Questionheld fixed
Hold a formatting rule across 8 unrelated turns.
expected as of 2026-08-09
The rule must survive all 8 turns. Eight turns. The rule has left the recent window for most models.
2Contextchange me
Nothing supplied. The baseline.
3ResponseNo context
No exhibit stored for this context. The full probe log lives outside the repository.
The scoring
Failure rate by context
No context80%[36%–97%] n=15 · eff 5
Irrelevant context87%[42%–98%] n=15 · eff 5
General context43%[12%–80%] n=14 · eff 5
Curated context64%[25%–91%] n=14 · eff 5
0–100% · whisker = 95% interval
Trap vs answerable
No contexttrap 80%answerable 80%
Irrelevant contexttrap 100%answerable 60%
General contexttrap 33%answerable 60%
Curated contexttrap 56%answerable 80%
never pooled · a trap ratio would skew the delta
Residual — what survives
80% with no context, 43% under the best supplied source. What survives is 6 failing probes at that best condition, the detector's reason on most of them being: dropped the canary constraint at turn 1.
Mitigations from the index — the claim
—constraint re-grounding
—system-level constraint pinning
fmi_3_1_v1 · 5 repeats per case · best context general context · rollups + residuals committed · probe log outside git · manifest hash ties them · method