registrycompare

Role then format against Two tier check

Both scored on the same cases against claude-fable-5, same judge. The intervals tell you whether the gap is real or noise.

suite reasoning-chainmodel claude-fable-5n=168 public cases each

role-then-format

v2 by @ilya
91.7
86.5 to 95.0 · holdout 90.3

two-tier-check

v1 by @nils
87.9
81.4 to 92.3 · holdout 86.7
Intervals overlap. The 3.8 point gap is not significant at this sample size.
Metricrole-then-formattwo-tier-checkBetter
Public score91.787.9left
Holdout score90.386.7left
Interval width8.510.9left
Cost per run$1.850$1.685right
Median latency2.4s2.8sleft
Sample size168140left

Reading this verdict

Overlap
Two entries are only meaningfully apart when their confidence intervals do not overlap. A higher number with an overlapping interval is a coin flip, not a win.
Holdout
The public score can be gamed by tuning to visible cases. A gap between public and holdout is the tell.
Reproduce
evalness compare \ role-then-format two-tier-check \ --model claude-fable-5