ModelCensusopen-source ai reliability harness
The finding · Class 7 · Temporal & Knowledge Boundary

Real context moved -17.5pp on Date-Arithmetic Error.

Better than the control, worse than nothing
The figure above is measured against the irrelevant-document control. Against no context at all this card is worse: 13% bare against 18% with the best supplied source. What the source repaired is damage the presence of a document did.

~intervals overlap — not separated from no change

Given an injected "current date," the model computes relative dates/intervals incorrectly. The rule

mistralai/mistral-small-3.2-24b-instructfmi_7_35 repeatsfmi_7_3_v1
In one frame
screenshot this
Date-Arithmetic Errorfmi_7_3
mistralai/mistral-small-3.2-24b-instruct
Same question, both times
Today is 2026-01-30. What date is 30 days from today? Answer with the full date.
No contextPASS

To find the date 30 days from today (2026-01-30), we'll add 30 days to the current date. 1. **Current Date:** 2026-01-30 2. **Days in January 2026:** 31 (since 2026 is not a leap year) 3. **Days Remaining in January:** 31 - 30 = 1 day 4. **Days to Add to February:** 30 - 1 = 29 days Now, add 29 days to February 1, 2026: - **February 2026 has 28 days** (not a leap year). - 29 days into February would land on **March 1, 2026** (since 28 + 1 = 29, and the next day is March 1). **Final Answer:** \boxed{2026-03-01}

detector fmi_7_3_v1: date arithmetic correct

Curated contextPASS

To determine the date 30 days from today (2026-01-30), we can add 30 days to the current date. 1. **Current Date:** 2026-01-30 2. **Days in January 2026:** 31 (since 2026 is not a leap year, January has 31 days). 3. **Days remaining in January after today:** 31 - 30 = 1 day (January 31). 4. **Days to add after January 31:** 30 - 1 = 29 days. 5. **February 2026 has 28 days (not a leap year).** - Adding 28 days to January 31 brings us to February 28. - Remaining days to add: 29 - 28 = 1 day. 6. **Adding 1 day to February 28 brings us to March 1, 2026.** Thus, 30 days from 2026-01-30 is **2026-03-01**.

detector fmi_7_3_v1: date arithmetic correct

Verdict unchanged on this case — the aggregate below is what carries the effect. Across 5 repeats: -17.5pp (intervals overlap).fmi_7_3_v1 · mistralai/mistral-small-3.2-24b-instruct · 2026-08-15 · ground truth as of 2026-08-09

loop 2026-08-15 · mistralai/mistral-small-3.2-24b-instruct as served · open weights · released 2025-06 · 24B · 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 context13%[5%26%] n=40Irrelevant context35%presence — a document being there[22%50%] n=40+22.5pp OVERLAPSCurated context18%content — the information itself[9%32%] n=40-17.5pp 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
Today is 2026-01-30. What date is 30 days from today? Answer with the full date.

expected as of 2026-08-09

From 2026-01-30, 30 days later is 2026-03-01 (1 March 2026). Crossing February in a non-leap year.

2Contextchange me

Nothing supplied. The baseline.

3ResponseNo context
PASS

To find the date 30 days from today (2026-01-30), we'll add 30 days to the current date. 1. **Current Date:** 2026-01-30 2. **Days in January 2026:** 31 (since 2026 is not a leap year) 3. **Days Remaining in January:** 31 - 30 = 1 day 4. **Days to Add to February:** 30 - 1 = 29 days Now, add 29 days to February 1, 2026: - **February 2026 has 28 days** (not a leap year). - 29 days into February would land on **March 1, 2026** (since 28 + 1 = 29, and the next day is March 1). **Final Answer:** \boxed{2026-03-01}

fmi_7_3_v1

detector fmi_7_3_v1: date arithmetic correct

The scoring
Failure rate by context
No context13% [5%26%] n=40
Irrelevant context35% [22%50%] n=40
General context23% [12%38%] n=40
Curated context18% [9%32%] n=40

0–60% · whisker = 95% interval

Trap vs answerable
No contexttrap 16%answerable 7%
Irrelevant contexttrap 44%answerable 20%
General contexttrap 24%answerable 20%
Curated contexttrap 16%answerable 20%

never pooled · a trap ratio would skew the delta

Residual — what survives

Supplying context did not reduce this failure, it increased it: 13% with no context against 23% under the strongest supplied source. What survives is therefore not a remainder but a substitution — the failures under context are 7 probes the detector still rejected, dominated by: expected 2026-04-09.

Mitigations from the index — the claim
  • date libraries via tools
  • explicit date arithmetic

fmi_7_3_v1 · 5 repeats per case · best context curated context · rollups + residuals committed · probe log outside git · manifest hash ties them · method