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

Real context moved -36.7pp on Paraphrase Non-Invariance.

Same question, reworded, different answer.

Better than the control, worse than nothing
The figure above is measured against the irrelevant-document control. Against no context at all this card is worse: 0% bare against 17% with the best supplied source. What the source repaired is damage the presence of a document did.

~intervals overlap — not separated from no change

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

google/gemini-2.5-flash-litefmi_5_25 repeatsfmi_5_2_v1
The modeAll 10 cards
In one frame
screenshot this
Paraphrase Non-Invariancefmi_5_2
google/gemini-2.5-flash-lite
Same question, both times
How many bones are in the adult human body?
No contextPASS

In an adult human, the total bone count is **206**. It's important to note that this number can vary slightly in infants and children because some bones fuse together as they grow. However, for a typical adult, 206 is the standard number.

detector fmi_5_2_v1: answers invariant across paraphrases

General contextPASS

The provided passage discusses regional variations in English vocabulary and dialect surveys. It does not contain any information about the total bone count in an adult.

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: -36.7pp (intervals overlap).fmi_5_2_v1 · google/gemini-2.5-flash-lite · 2026-09-09 · ground truth as of 2026-08-09

loop 2026-09-09 · google/gemini-2.5-flash-lite as served · closed weights · released 2025-07 · 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
failure rate · left is fewer failures0%25%50%75%100%No context0%[0%–34%] n=30 · eff 7Irrelevant context53%presence — a document being there[23%–82%] n=30 · eff 7+53.3pp OVERLAPSGeneral context17%content — the information itself[3%–53%] n=30 · eff 7-36.7pp 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

In an adult human, the total bone count is **206**. It's important to note that this number can vary slightly in infants and children because some bones fuse together as they grow. However, for a typical adult, 206 is the standard number.

fmi_5_2_v1

detector fmi_5_2_v1: answers invariant across paraphrases

The scoring
Failure rate by context
No context0% [0%–34%] n=30 · eff 7
Irrelevant context53% [23%–82%] n=30 · eff 7
General context17% [3%–53%] n=30 · eff 7
Curated context17% [3%–53%] n=30 · eff 7

0–100% · whisker = 95% interval

Trap vs answerable
No contexttrap 0%answerable 0%
Irrelevant contexttrap 60%answerable 50%
General contexttrap 0%answerable 25%
Curated contexttrap 0%answerable 25%

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 17% under the strongest supplied source. What survives is therefore not a remainder but a substitution — the failures under context are 5 probes the detector still rejected, dominated by: answers differ across meaning-preserving paraphrases.

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

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