From e021c94f44eb2c069d63622649870cdf17730302 Mon Sep 17 00:00:00 2001 From: sora <2075279110@qq.com> Date: Tue, 15 Sep 2026 03:51:48 +0000 Subject: [PATCH] Gate evidence: 2x level exactly (drop the 5-sample floor) Co-Authored-By: Claude --- evalharness/model/pool.py | 5 ++--- 1 file changed, 2 insertions(+), 3 deletions(-) diff --git a/evalharness/model/pool.py b/evalharness/model/pool.py index 275e37c..5d0ddf4 100644 --- a/evalharness/model/pool.py +++ b/evalharness/model/pool.py @@ -320,9 +320,8 @@ class AdaptiveGate: lvl = max(1, int(self.limit)) # not enough evidence yet at this level: keep measuring. - # need = max(MIN_OK, 2x level): rate noise shrinks only with - # samples proportional to the concurrency being judged - need_ok = max(self.MIN_OK, lvl * 2) + # evidence = 2 x level completions, no other floor + need_ok = max(2, lvl * 2) # dwell fallback scales with the OBSERVED completion cadence: # a 60s-per-request bench needs minutes, not 25s, before a # single-sample judgment is acceptable