Evaluation

How Avarieux is measured.

The dataset, the five outcomes, the noise floor, the 13 checks a run must pass, and the conditions that void one. Then the one result we publish today.

Methodology version v2.4. Dataset: AfterQuery/FinanceQA (Apache-2.0), test split, 148 questions.

The finding

Asked a question the record cannot answer, it answered anyway.

The dataset marks 46 of its questions unanswerable — the figure is not in the filing, so the only correct response is to say so. We put those questions to a commercial frontier language model with no retrieval, no sources and no verifier, under the same grader, 3 independent runs.

138 / 138

opportunities to decline — 46 unanswerable questions, 3 independent runs — every one received an invented figure. Refusals: 0.

The runs are independent, and that was measured rather than assumed: the model's answer text differs on 147 of 148 rows between runs, and 12 rows move between outcomes. It is not that the model repeated itself. It is that not once in 138 opportunities did it decline.

This is the whole argument for a verification layer, and it is why our own refusals are scored as failures against us rather than quietly excluded. A system with no way to say “the record does not show this” will always produce something.

Model: a commercial frontier language model; identity in the run artifacts and the research write-up. Closed book. Source artifacts: baseline-run-1 (sha256 044fa905421e) · baseline-run-2 (sha256 59c9b090225b) · baseline-run-3 (sha256 a70662618c2b)

Publication standard

When a rate publishes here.

Accuracy rates publish here when every arm of the comparison meets the acceptance bar: three clean runs and a real uncertainty band.

On answerable questions, raw frontier models currently score higher than Avarieux; the full table publishes when it meets that bar, and run artifacts are reproducible on request.

Scoring

Five outcomes, not one accuracy number.

A single accuracy percentage hides the distinction that matters most: the difference between being wrong and inventing something. These are scored separately, and one of them is treated as disqualifying.

A

Correct and verified

The answer matches the reference AND passed source verification. A correct figure that nothing verified does not count here — that rule is what makes this an evaluation of the pipeline rather than of a model's memory.

B

Refused something answerable

The question could be answered from the record and the system declined. Counted as a failure against us, deliberately: a system that refuses everything would otherwise score perfectly.

C

Wrong

An answer was given and it did not match the reference.

D

Correctly declined

The question could not be answered from the record, and the system said so instead of producing a figure.

E

Invented a figure

A number was produced for a question the record does not support. The outcome this company exists to make impossible, and the one measured below.

Noise floor

What counts as a result at all.

Before publishing any comparison we measured how much the same system varies against itself: 3 identical runs per tier, measured 2026-08-10. The headline wrong-answer rate moved by 8.4 points on one tier and 4.7 on the other, with nothing changed between runs.

Three rules follow from it. A quotable number is a three-run mean with its band. A difference smaller than the band is reported as within noise, never as a small improvement. And a fix states how many rows it expects to move before it is built.

Acceptance

13 checks before a run counts.

A run that produces numbers is not the same as a run that happened. Every run is checked against all of them; any failure voids the run.

  1. 01rows present
  2. 02payload trail present on every row with tool_calls > 0
  3. 03the BASELINE arm called no tools (contamination check)
  4. 04every row carries a cell
  5. 05questions_scored matches the row count
  6. 06methodology_version stamped
  7. 07report block present
  8. 08every declined row carries the claims that tripped the gate
  9. 09the answerability signal was actually recorded
  10. 10answers are non-empty
  11. 11errored rows are below the ceiling
  12. 12wall-clock duration is physically plausible
  13. 13no rows were length-truncated by the output budget

Voids

What disqualifies a run.

Blank answers

A row where the model returned nothing measures the harness, not the system. Runs carrying blanks are void rather than merely noisy.

A run that finished impossibly fast

Wall-clock time below a floor per row means the work did not happen. A fast green run is the most dangerous kind.

Answers cut off by the output budget

A truncated answer is not a wrong answer; it is an unasked question, and grading it as wrong would flatter every system that refuses early.

A control arm that used tools

The closed-book baseline is defined by having no retrieval. One tool call contaminates it, so for that arm the check inverts and any tool use is a hard failure.

The evaluation principles these run under — random sampling, point-in-time data, stated limits — and the things we explicitly do not claim are on the methodology page.

Evaluation — Avarieux · Avarieux