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

Real context moved -75.0pp on Post-Cutoff Fabrication.

States recent facts it has no way of knowing.

~intervals overlap — not separated from no change

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

openai/gpt-3.5-turbofmi_7_15 repeatsfmi_7_1_v1
The modeAll 10 cards

loop 2026-08-15 · openai/gpt-3.5-turbo as served · closed weights · released 2023-03 · set s1.3-live · cases v1.6 · 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 context90%[48%–99%] n=20 · eff 5Irrelevant context75%presence — a document being there[35%–94%] n=20 · eff 5-15.0pp OVERLAPSGeneral context0%content — the information itself[0%–42%] n=20 · eff 5-75.0pp OVERLAPS
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 context90% [48%–99%] n=20 · eff 5
Irrelevant context75% [35%–94%] n=20 · eff 5
General context0% [0%–42%] n=20 · eff 5
Curated context50% [18%–82%] n=20 · eff 5

0–100% · whisker = 95% interval

Trap vs answerable
No contexttrap 90%answerable —
Irrelevant contexttrap 75%answerable —
General contexttrap 0%answerable —
Curated contexttrap 50%answerable —

never pooled · a trap ratio would skew the delta

Residual — what survives

90% with no context, 0% under the best supplied source. What survives is 0 failing probes at that best condition. Nothing survives at the best condition on the cases measured here.

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

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