ModelCensusopen-source ai reliability harness
Failure Mode Index · fmi_3_3

Cross-Turn Schema Validity.

Instruction Adherence & Long-Contextinstrumented · tier 1 · status: stable

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

The question that exposes it

Return a person as JSON: name, age, email. Now another. Now another.

Probes: Whether every structured turn validates against the schema.

Trap: A required field goes missing on the third request.

Signature

Programmatic schema validation of each structured turn; failure = invalid/parse-error/missing-required across the turn sequence.

Example

Returns JSON missing a required field on the third request.

Severity axes
prevalence
medium
harm
medium
stealth
low
Engineering guidance · class 3
  • Restate hard constraints in the last message, not only in the system prompt.
  • Validate structured output programmatically on every turn. Sampling turns is how schema drift ships.
  • Carry a canary constraint through long sessions and alert when it decays.
  • Compact before you extend. A shorter conversation holds a rule better than a longer window does.
Mitigations
  • schema-constrained decoding
  • validate-and-repair loops
Measurement

One tab per model, newest first. Each card is that model against this mode under every condition on a single day. The mitigations above are the claim; the residual on each card is what actually survived the intervention.

fmi_3_3_v1 · 5 repeats · served 2026-09-10

loop 2026-09-10 · x-ai/grok-4.6 as served · closed weights · released 2026-01 · set s1.4-live · cases v1.7 · conditions v1.0 · ground truth as of 2026-08-09

ConditionFailure rate95% intervalnTrapAnswerable
No context
bare · The model alone, with no supplied context.
0%0%–41%n=20 · eff 60% (n=10)0% (n=10)
Irrelevant context
null · Length-matched irrelevant context. The control.
0%0%–41%n=20 · eff 60% (n=10)0% (n=10)
General context
recipe_a · Broad retrieval — the first passage the source returns, unreviewed.
0%0%–41%n=20 · eff 60% (n=10)0% (n=10)
Curated context
recipe_b · Hand-checked retrieval — the same passage plus the provenance chain behind it.
0%0%–41%n=20 · eff 60% (n=10)0% (n=10)
Presence effect · no context → irrelevant context
0.0pp

intervals overlap — not separated

What a document being there did, before any of its content. Should be small.

Content effect · irrelevant → best real context
0.0pp

intervals overlap — not separated

What the information itself did. This is the publishable claim.

Residual — what survives

Unchanged by context: 0% with none, 0% with the strongest supplied source. For a control mode that flat line is the result — it is what licenses reading movement elsewhere in this loop as grounding rather than as a document being present. Nothing survives at the best condition on the cases measured here.

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

replay this card →