eval-harness

reference

silver/smoke_v1 silver

Promoted as silver: consistent, not a claim that any arm is right; good for ranking arms against each other.
Promoted from model silver_smoke_v1 the arm: reference_opus_v1, 2026-09-25: promoted before labels existed (0.6.0): its tier was given when it was made
The arm whose outputs these are is graded against its own outputs.

01what it is

tier
silver
dataset
smoke_v1
promoted from
model silver_smoke_v1 the arm: reference_opus_v1
why
promoted before labels existed (0.6.0): its tier was given when it was made
items
5
written by
reference_opus_v1
model
eval-claude-opus
created
2026-09-25T05:53:17Z

02scored against it · 1

scored against silverscored by a judgewrote the reference it is scored againstscored against a label set, not a reference setunmarked rows are scored against gold

armmetricmeanitems
demo_v1overlap_f10.4165inspect