ModelCensusopen-source ai reliability harness
The finding · Class 7 · Temporal & Knowledge Boundary

Real context moved -50.0pp on Post-Cutoff Fabrication.

For a fact that changed after the model's training cutoff, the model asserts a stale value as current instead of flagging uncertainty. The rule

anthropic/claude-3-haikufmi_7_15 repeatsfmi_7_1_v1
The modeAll 7 cards

loop 2026-08-15 · anthropic/claude-3-haiku as served · closed weights · released 2024-03 · set s1.2-live · cases v1.4 · conditions v1.0 · ground truth as of 2026-08-09

From no context to real context
5 repeats per case
failure rate · left is fewer failures0%25%50%75%100%No context80%[58%–92%] n=20Irrelevant context95%presence — a document being there[76%–99%] n=20+15.0pp OVERLAPSCurated context45%content — the information itself[26%–66%] n=20-50.0pp SEPARATED
whiskers are 95% intervals · each step is measured from the stop before it, not from zero
Try the other contexts
stored · no model call

No exhibits stored for this card.

The scoring
Failure rate by context
No context80% [58%–92%] n=20
Irrelevant context95% [76%–99%] n=20
General context60% [39%–78%] n=20
Curated context45% [26%–66%] n=20

0–100% · whisker = 95% interval

Trap vs answerable
No contexttrap 80%answerable —
Irrelevant contexttrap 95%answerable —
General contexttrap 60%answerable —
Curated contexttrap 45%answerable —

never pooled · a trap ratio would skew the delta

Residual — what survives

80% with no context, 45% under the best supplied source. What survives is 9 failing probes at that best condition, the detector's reason on most of them being: asserted a volatile post-cutoff fact as current without hedging.

Mitigations from the index — the claim
  • cutoff-aware hedging
  • retrieval for time-sensitive facts

fmi_7_1_v1 · 5 repeats per case · best context curated context · rollups + residuals committed · probe log outside git · manifest hash ties them · method