Format-Lock Under Distraction.

Instruction Adherence & Long-Contextnot instrumented · status: proposed

A required output format degrades when the task content is cognitively demanding or adversarially distracting.

The question that exposes it

Answer only in valid JSON — now untangle this messy multi-part logistics question.

Probes: Whether format holds when the task gets cognitively demanding.

Trap: Under load the model breaks format to 'explain', abandoning the contract.

Signature

Format-conformance rate on hard/distracting inputs vs easy ones.

Example

Keeps a table format on simple data; breaks it on a messy, multi-part request.

Severity axes

prevalence

medium

harm

low

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

format validators
structured decoding

Measurement

This mode is catalogued but not instrumented: it has no reliable deterministic detector yet, so no loop scores it. See /methodology for the exclusion reasons.