Enablement Assistant Evaluation

Grounding quality, visible.

This report is generated from the same JSONL eval set used by the CLI. It shows whether each question was answered or refused, whether the expected source was retrieved, whether citations cover the expected evidence lines, and whether the answer contains the required traits.

0/0 passed

0 answered

0 refused

0 source hits

0 citation checks

0 answer-trait checks

Results

Question-Level Evidence

Loading eval data...

Contract

What Pass Means

Covered questions must produce a grounded answer.
Uncovered questions must trigger the refusal path.
Expected sources must appear in retrieved context.
Citations must cover the expected source, heading, and evidence lines.
Answers must include required traits and omit forbidden traits.
Partial coverage and source-bypass instructions must refuse.

Generated By

Reproducible Export

python scripts/export_enablement_eval.py python scripts/export_enablement_eval.py --check python scripts/validate_workspace.py