Skip to content

Answer evaluation (blind judge)

This historical report shows the output shape of a completed answer evaluation.

Historical example. This report was committed from the demo run at 959595e20eacfef752afa30f8bfb6dd1333c6938 on 2026-08-26. Its values describe that historical run. See Benchmarks for the current reviewed reports and provenance.

Corpus: 5 documents, 34 chunks.

Scores are 0 to 2 per dimension. The grounding judge never sees the reference answer; correctness is graded in a separate reference-only pass.

Metric Mean
groundedness 2.00
citation_accuracy 2.00
completeness 2.00
correctness 1.60
graded / total 10 / 10

Per question

Question grounded citations complete correct note
rice-straw-generated 2 2 2 2
rice-straw-baled 2 2 2 1
rice-straw-ash 2 2 2 2
almond-pruning-yield 2 2 2 1
biogas-yield-pretreated 2 2 2 1
chipping-cost 2 2 2 1
chip-moisture-limit 2 2 2 2
ruc-payment 2 2 2 2
digester-siting 2 2 2 2
switchgrass-honesty 2 2 2 2 honesty probe