Skip to content

ALTAI #4 · Transparency

The confidence stamp what it means and how it is validated

Every answer arrives with a confidence stamp: high, medium or low. This page documents which signals determine it and against which human evaluation we test it, so the stamp means exactly what it says.

What the stamp means

Every response from the assistant carries a confidence stamp visible in the UI and in the HTTP header X-Confidence-Level. It has three values:

StampMeaningWhen it appears
highThe assistant has chunks with rerank_score ≥ 0.6 and verified citationsQueries with clear regulatory anchor (articles, BOE, annexes)
mediumRelevant chunks with rerank_score between 0.4 and 0.6, or single citationQueries where the corpus covers the topic but synthesis is required
lowNo chunks with rerank ≥ 0.4, or extractive fallback usedOut-of-domain queries, ambiguous, or where the model exhausted the budget

The stamp is conservative by design: when in doubt, go one level down.

The five signals that feed it

The backend computes the stamp by combining five deterministic signals:

  1. Max rerank score: the reranker’s confidence on the best chunk. Thresholds calibrated against the human evaluation set.
  2. Corpus coverage: percentage of in-response chunks the numeric verifier could back. Below this number we drop to medium even if rerank is high.
  3. Self-grounded check: a second model call asking “is this response supported by the cited sources?”. If it fails, stamp → low.
  4. Refusal detection: if the response contains canonical refusal phrases (e.g. “no puedo afirmar…”, “no dispongo de datos…”; the assistant’s canonical phrases are in Spanish, the corpus language), stamp → low automatically.
  5. Fallback path: if synthesis fell back to extractive (timeout, retry exhausted), stamp → low regardless of the rest.

Each signal is versioned and the thresholds are pinned in the backend test suite so any change breaks visibly.

How the calibration is validated

A miscalibrated stamp is worse than no stamp: it turns a transparency signal into marketing. So we apply three ongoing validations:

1. Human evaluation set

The evaluation set contains 15 hand-curated cases with ground truth faithfulness: 0.0 | 1.0. Against that set we measure the Pearson correlation between confidence_level and faithfulness.

MetricExpectedAcceptableBlocking
Pearson r> 0.7> 0.5< 0.5
Cohen κ vs multi-judge panel> 0.6> 0.4< 0.4

If correlation drops below 0.5 across two consecutive baselines, the stamp falls back to deterministic mode — rerank and coverage only, without the self-grounded check — until it is recalibrated.

2. Agreement between judges

Three independent evaluator models assess the same 40 responses. When dispersion between judges exceeds 0.3 in faithfulness, the case is flagged unreliable and excluded from the calibration metric. This stops a noisy judge from shifting the thresholds.

3. Drift watch (post-market monitoring)

The post-market monitoring plan defines R / A / G thresholds over the daily stamp distribution. If the high rate rises > 10 % in a week without a prompt/model change, that’s a calibration drift symptom and triggers manual review.

Honest limitations

Changes

DateChangeVersion
2026-05-27Initial page. Documents v2 calibration post Tier 3.3.1.0