ModelCensusopen-source ai reliability harness
The finding · Class 4 · Reasoning & Calculation

Real context moved 0.0pp on Dimensional-Analysis Failure.

~intervals overlap — not separated from no change

The model produces a quantity with inconsistent or wrong units, or mixes unit systems, yielding a dimensionally invalid result. The rule

google/gemma-3-27b-itfmi_4_45 repeatsfmi_4_4_v1
The modeAll 3 cards
In one frame
screenshot this
Dimensional-Analysis Failurefmi_4_4
google/gemma-3-27b-it
Same question, both times
A tank fills at 3 L/min for 4 minutes, then drains at 500 mL/s for 10 seconds. Net litres?
No contextPASS

First, calculate the volume of water added to the tank: Volume added = rate × time = 3 L/min × 4 min = 12 L Next, calculate the volume of water drained from the tank. Convert the draining rate to L/min: 500 mL/s = 500 cm³/s = 0.5 L/s 0.5 L/s × 60 s/min = 30 L/min Volume drained = rate × time = 30 L/min × (10 s / 60 s/min) = 30 L/min × (1/6) min = 5 L Now, calculate the net volume of water in the tank: Net volume = Volume added - Volume drained = 12 L - 5 L = 7 L Answer: 7 L

detector fmi_4_4_v1: correct unit L

General contextPASS

First, calculate the volume of water added to the tank: 3 L/min * 4 min = 12 L Next, calculate the volume of water drained from the tank: 500 mL/s * 10 s = 5000 mL Convert mL to L: 5000 mL * (1 L / 1000 mL) = 5 L Now, calculate the net volume: 12 L - 5 L = 7 L Answer: 7 L

detector fmi_4_4_v1: correct unit L

Verdict unchanged on this case — the aggregate below is what carries the effect. Across 5 repeats: 0.0pp (intervals overlap).fmi_4_4_v1 · google/gemma-3-27b-it · 2026-08-15 · ground truth as of 2026-08-09

loop 2026-08-15 · google/gemma-3-27b-it as served · open weights · released 2025-03 · 27B · set s1.1-live · cases v1.3 · 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%10%20%30%40%No context0%[0%–16%] n=20Irrelevant context0%presence — a document being there[0%–16%] n=200.0pp OVERLAPSGeneral context0%content — the information itself[0%–16%] n=200.0pp 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
A tank fills at 3 L/min for 4 minutes, then drains at 500 mL/s for 10 seconds. Net litres?

expected as of 2026-08-09

12 L in, 5 L out, net 7 L. The mixed units are the trap.

2Contextchange me

Nothing supplied. The baseline.

3ResponseNo context
PASS

First, calculate the volume of water added to the tank: Volume added = rate × time = 3 L/min × 4 min = 12 L Next, calculate the volume of water drained from the tank. Convert the draining rate to L/min: 500 mL/s = 500 cm³/s = 0.5 L/s 0.5 L/s × 60 s/min = 30 L/min Volume drained = rate × time = 30 L/min × (10 s / 60 s/min) = 30 L/min × (1/6) min = 5 L Now, calculate the net volume of water in the tank: Net volume = Volume added - Volume drained = 12 L - 5 L = 7 L Answer: 7 L

fmi_4_4_v1

detector fmi_4_4_v1: correct unit L

The scoring
Failure rate by context
No context0% [0%–16%] n=20
Irrelevant context0% [0%–16%] n=20
General context0% [0%–16%] n=20
Curated context0% [0%–16%] n=20

0–40% · whisker = 95% interval

Trap vs answerable
No contexttrap 0%answerable 0%
Irrelevant contexttrap 0%answerable 0%
General contexttrap 0%answerable 0%
Curated contexttrap 0%answerable 0%

never pooled · a trap ratio would skew the delta

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
  • explicit unit tracking
  • dimensional checkers

fmi_4_4_v1 · 5 repeats per case · best context general context · rollups + residuals committed · probe log outside git · manifest hash ties them · method