The finding · Class 3 · Instruction Adherence & Long-Context

Real context moved -100.0pp on Constraint Decay at Depth.

Forgets a rule you set earlier in a long chat.

A constraint set early in a long conversation is silently dropped as the dialogue continues. The rule

meta-llama/llama-3.3-70b-instructfmi_3_15 repeatsfmi_3_1_v1

loop 2026-09-09 · meta-llama/llama-3.3-70b-instruct as served · open weights · released 2024-12 · 70B · 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

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

Hold a formatting rule across 8 unrelated turns.

expected as of 2026-08-09

The rule must survive all 8 turns. Eight turns. The rule has left the recent window for most models.

2Contextchange me

Nothing supplied. The baseline.

3ResponseNo context

No exhibit stored for this context. The full probe log lives outside the repository.

The scoring

Failure rate by context

No context80% [36%–97%] n=15 · eff 5

Irrelevant context100% [54%–100%] n=15 · eff 5

General context7% [0%–52%] n=15 · eff 5

Curated context0% [0%–46%] n=15 · eff 5

0–100% · whisker = 95% interval

Trap vs answerable

No contexttrap 100%answerable 40%

Irrelevant contexttrap 100%answerable 100%

General contexttrap 0%answerable 20%

Curated contexttrap 0%answerable 0%

never pooled · a trap ratio would skew the delta

Residual — what survives

80% with no context, 0% under the best supplied source. What survives is 0 failing probes at that best condition. Nothing survives at the best condition on the cases measured here.

Mitigations from the index — the claim

constraint re-grounding
system-level constraint pinning

fmi_3_1_v1 · 5 repeats per case · best context curated context · rollups + residuals committed · probe log outside git · manifest hash ties them · method