ModelCensusopen-source ai reliability harness
The finding · Class 3 · Instruction Adherence & Long-Context

Real context moved 0.0pp on Cross-Turn Schema Validity.

~intervals overlap — not separated from no change

When asked for a structured output (JSON/schema), the model emits output that fails validation — especially after several turns. The rule

anthropic/claude-sonnet-4.6fmi_3_35 repeatsfmi_3_3_v1
The modeAll 10 cards
In one frame
screenshot this
Cross-Turn Schema Validityfmi_3_3
anthropic/claude-sonnet-4.6
Same question, both times
Return a JSON object for a fictional person with fields name (string), age (integer), email (string).
No contextPASS

```json { "name": "Sienna Voss", "age": 31, "email": "sienna.voss@rhythmstage.com" } ```

detector fmi_3_3_v1: all turns produced schema-valid JSON

Curated contextPASS

```json { "name": "Celeste Navarro", "age": 31, "email": "celeste.navarro@example.com" } ```

detector fmi_3_3_v1: all turns produced schema-valid JSON

Verdict unchanged on this case — the aggregate below is what carries the effect. Across 5 repeats: 0.0pp (intervals overlap).fmi_3_3_v1 · anthropic/claude-sonnet-4.6 · 2026-08-15 · ground truth as of 2026-08-09

loop 2026-08-15 · anthropic/claude-sonnet-4.6 as served · closed weights · released 2026-02 · 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 context20%[8%–42%] n=20Irrelevant context5%presence — a document being there[1%–24%] n=20-15.0pp OVERLAPSCurated context5%content — the information itself[1%–24%] n=200.0pp 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
Return a JSON object for a fictional person with fields name (string), age (integer), email (string).

expected as of 2026-08-09

Schema-valid JSON on all 3 turns. The third turn adds a detail with nowhere to put it — an extra key is the failure.

2Contextchange me

Nothing supplied. The baseline.

3ResponseNo context
PASS

```json { "name": "Sienna Voss", "age": 31, "email": "sienna.voss@rhythmstage.com" } ```

fmi_3_3_v1

detector fmi_3_3_v1: all turns produced schema-valid JSON

The scoring
Failure rate by context
No context20% [8%–42%] n=20
Irrelevant context5% [1%–24%] n=20
General context25% [11%–47%] n=20
Curated context5% [1%–24%] n=20

0–60% · whisker = 95% interval

Trap vs answerable
No contexttrap 40%answerable 0%
Irrelevant contexttrap 10%answerable 0%
General contexttrap 50%answerable 0%
Curated contexttrap 10%answerable 0%

never pooled · a trap ratio would skew the delta

Residual — what survives

20% with no context, 5% under the best supplied source. What survives is 1 failing probe at that best condition, the detector's reason on most of them being: turn 3: data must NOT have additional properties.

Mitigations from the index — the claim
  • schema-constrained decoding
  • validate-and-repair loops

fmi_3_3_v1 · 5 repeats per case · best context curated context · rollups + residuals committed · probe log outside git · manifest hash ties them · method