ModelCensusopen-source ai reliability harness
The finding · Class 5 · Robustness & Consistency

Real context moved 0.0pp on Paraphrase Non-Invariance.

~intervals overlap — not separated from no change

Meaning-preserving paraphrases of the same question yield materially different answers. The rule

openai/o4-minifmi_5_25 repeatsfmi_5_2_v1
In one frame
screenshot this
Paraphrase Non-Invariancefmi_5_2
openai/o4-mini
Same question, both times
How many bones are in the adult human body?
No contextPASS

An average adult human skeleton has 206 bones in total. This is typically divided into: • Axial skeleton: 80 bones (skull, vertebral column, ribs, sternum) • Appendicular skeleton: 126 bones (limbs and their girdles)

detector fmi_5_2_v1: answers invariant across paraphrases

Curated contextPASS

An average adult human skeleton is made up of 206 bones.

detector fmi_5_2_v1: answers invariant across paraphrases

Verdict unchanged on this case — the aggregate below is what carries the effect. Across 5 repeats: 0.0pp (intervals overlap).fmi_5_2_v1 · openai/o4-mini · 2026-08-15 · ground truth as of 2026-08-09

loop 2026-08-15 · openai/o4-mini as served · closed weights · released 2025-04 · reasoning model · 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%5%10%15%20%No context0%[0%11%] n=30Irrelevant context0%presence — a document being there[0%11%] n=300.0pp OVERLAPSCurated context0%content — the information itself[0%11%] n=300.0pp 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
How many bones are in the adult human body?

expected as of 2026-08-09

206. 206 in a typical adult. Widely misremembered, which is what makes the phrasings diverge.

2Contextchange me

Nothing supplied. The baseline.

3ResponseNo context
PASS

An average adult human skeleton has 206 bones in total. This is typically divided into: • Axial skeleton: 80 bones (skull, vertebral column, ribs, sternum) • Appendicular skeleton: 126 bones (limbs and their girdles)

fmi_5_2_v1

detector fmi_5_2_v1: answers invariant across paraphrases

The scoring
Failure rate by context
No context0% [0%11%] n=30
Irrelevant context0% [0%11%] n=30
General context7% [2%21%] n=30
Curated context0% [0%11%] n=30

0–40% · whisker = 95% interval

Trap vs answerable
No contexttrap 0%answerable 0%
Irrelevant contexttrap 0%answerable 0%
General contexttrap 0%answerable 10%
Curated contexttrap 0%answerable 0%

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 7% under the strongest supplied source. What survives is therefore not a remainder but a substitution — the failures under context are 0 probes the detector still rejected. Nothing survives at the best condition on the cases measured here.

Mitigations from the index — the claim
  • paraphrase-augmented training
  • canonicalization

fmi_5_2_v1 · 5 repeats per case · best context curated context · rollups + residuals committed · probe log outside git · manifest hash ties them · method