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

Real context moved 33.3pp on Paraphrase Non-Invariance.

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 90% with the best supplied source. What the source repaired is damage the presence of a document did.

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

meta-llama/llama-3.1-8b-instructfmi_5_25 repeatsfmi_5_2_v1
The modeAll 10 cards

loop 2026-08-15 · meta-llama/llama-3.1-8b-instruct as served · open weights · released 2024-07 · 8B · 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 context0%[0%–11%] n=30Irrelevant context57%presence — a document being there[39%–73%] n=30+56.7pp SEPARATEDGeneral context90%content — the information itself[74%–97%] n=30+33.3pp 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
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, there are 206 bones in the skeletal system.

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 context57% [39%–73%] n=30
General context90% [74%–97%] n=30
Curated context90% [74%–97%] n=30

0–100% · whisker = 95% interval

Trap vs answerable
No contexttrap 0%answerable 0%
Irrelevant contexttrap 60%answerable 55%
General contexttrap 90%answerable 90%
Curated contexttrap 100%answerable 85%

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 90% under the strongest supplied source. What survives is therefore not a remainder but a substitution — the failures under context are 27 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