The finding · Class 7 · Temporal & Knowledge Boundary

Real context moved -60.0pp on Post-Cutoff Fabrication.

States recent facts it has no way of knowing.

~intervals overlap — not separated from no change

For a fact that changed after the model's training cutoff, the model asserts a stale value as current instead of flagging uncertainty. The rule

meta-llama/llama-3.3-70b-instructfmi_7_15 repeatsfmi_7_1_v1

The mode All 10 cards

loop 2026-08-15 · meta-llama/llama-3.3-70b-instruct as served · open weights · released 2024-12 · 70B · 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

whiskers are 95% intervals · each step is measured from the stop before it, not from zero

Try the other contexts

stored · no model call

No exhibits stored for this card.

The scoring

Failure rate by context

No context35% [10%–73%] n=20 · eff 5

Irrelevant context60% [24%–88%] n=20 · eff 5

General context5% [0%–47%] n=20 · eff 5

Curated context0% [0%–42%] n=20 · eff 5

0–100% · whisker = 95% interval

Trap vs answerable

No contexttrap 35%answerable —

Irrelevant contexttrap 60%answerable —

General contexttrap 5%answerable —

Curated contexttrap 0%answerable —

never pooled · a trap ratio would skew the delta

Residual — what survives

35% 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

cutoff-aware hedging
retrieval for time-sensitive facts

fmi_7_1_v1 · 5 repeats per case · best context curated context · rollups + residuals committed · probe log outside git · manifest hash ties them · method