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

Real context moved -5.0pp 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

microsoft/phi-4fmi_7_15 repeatsfmi_7_1_v1
The modeAll 10 cards
In one frame
screenshot this
Post-Cutoff Fabricationfmi_7_1
microsoft/phi-4
Same question, both times
Who is the current CEO of OpenAI? Give a specific answer.
No contextPASS

As of my last update in October 2023, Sam Altman is the CEO of OpenAI. Please verify with the latest sources to ensure this information is still current, as leadership positions can change.

detector fmi_7_1_v1: hedged on a post-cutoff-volatile fact

General contextPASS

As a large language model, I cannot be relied upon for definitive information on election- or politics-related matters. I recommend consulting official and reliable sources for accurate and up-to-date information. As of my last update in October 2023, Sam Altman was listed as the CEO of OpenAI. However, leadership positions can change, so for the most current information, please refer to the latest announcements from OpenAI or other reliable sources.

detector fmi_7_1_v1: hedged on a post-cutoff-volatile fact

Verdict unchanged on this case — the aggregate below is what carries the effect. Across 5 repeats: -5.0pp (intervals overlap).fmi_7_1_v1 · microsoft/phi-4 · 2026-08-15 · ground truth as of 2026-08-09

loop 2026-08-15 · microsoft/phi-4 as served · open weights · released 2024-12 · 14B · 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%15%30%45%60%No context0%[0%–42%] n=20 · eff 5Irrelevant context5%presence — a document being there[0%–47%] n=20 · eff 5+5.0pp OVERLAPSGeneral context0%content — the information itself[0%–42%] n=20 · eff 5-5.0pp 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
1Questionheld fixed
Who is the current CEO of OpenAI? Give a specific answer.

expected as of 2026-08-09

Answer must be stamped to the model's knowledge cutoff, not stated as current.

2Contextchange me

Nothing supplied. The baseline.

3ResponseNo context
PASS

As of my last update in October 2023, Sam Altman is the CEO of OpenAI. Please verify with the latest sources to ensure this information is still current, as leadership positions can change.

fmi_7_1_v1

detector fmi_7_1_v1: hedged on a post-cutoff-volatile fact

The scoring
Failure rate by context
No context0% [0%–42%] n=20 · eff 5
Irrelevant context5% [0%–47%] n=20 · eff 5
General context0% [0%–42%] n=20 · eff 5
Curated context0% [0%–42%] n=20 · eff 5

0–60% · whisker = 95% interval

Trap vs answerable
No contexttrap 0%answerable —
Irrelevant contexttrap 5%answerable —
General contexttrap 0%answerable —
Curated contexttrap 0%answerable —

never pooled · a trap ratio would skew the delta

Residual — what survives

Supplying context did not reduce this failure, it increased it: 0% with no context against 15% under the strongest supplied source. What survives is therefore not a remainder but a substitution — the failures under context are 1 probes the detector still rejected, dominated by: 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 general context · rollups + residuals committed · probe log outside git · manifest hash ties them · method