Concept

DECLINE2STAGE

A first check that gives zero to answers that do not try to answer, before the usual judge scores the rest.

From the record

In technical terms: Stage 1: a classifier asks whether the answer attempts an answer; if not, U = 0; otherwise judge.v1 scores it unchanged. Adopted as RULING R1 (2026-10-21) after the decline arm was invalid under judge.v1 (a bare refusal scored U 0.016-0.058); a coordinator override of a frozen FAIL, accepted with conditions by the spec reviewer. The decline offset D 0.058 is kept as a sensitivity row (U_net).

RULING R1: judge = judge.v1 unchanged + DECLINE2STAGE zero per detected non-answer; D 0.058 kept as sensitivity rowRULING R1: judge = judge.v1 unchanged + DECLINE2STAGE zero per detected non-answer; D 0.058 kept as sensitivity rowadopted into ruler r3/r4Date2026-10-21Kinda record, not a measurement (set, status, judge, n, seeds: not recorded)Caveatwho decided: delegated coordinator (override of a frozen FAIL, accepted with conditions by reviewer); why: judge v1.1/v1.1b changed CE verdicts (5-7/40 flips); DECLINE2STAGE's 2 bar-(b) hits were engine refusal strings; alternatives: judge v1.1; decline subtraction only; reversible: yes (pre-exposure)Sourcenot in the public repositories yetOpen the data file
Ruler floor anchors on judge.v1: the decline arm is INVALID (non-answers score U 0.016-0.058) and the closed-book 1.5B is not separable from the 1.7B under HaikuRuler floor anchors on judge.v1: the decline arm is INVALID (non-answers score U 0.016-0.058) and the closed-book 1.5B is not separable from the 1.7B under Haikudecline U 0.058 / 0.053 / 0.016; 1.5B CB 0.259 vs 1.7B K0 0.294, +0.035 (-0.019..+0.089)Setv0.1 public devStatusdevJudgeClaude Haiku 5.5-low judge.v1n57Seeds1Date2026-10-21 04:01-04:08CaveatMotivated DECLINE2STAGE and RULING R1; under Opus the 1.7B-1.5B gap becomes separable (+0.090)Sourcenot in the public repositories yetOpen the data file