registrycompare

Role then format against Source span citation

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

source-span-cite

v2 by @marek
93.8
87.7 to 96.9 · holdout 93.8
Intervals overlap. The 2.1 point gap is not significant at this sample size.
Metricrole-then-formatsource-span-citeBetter
Public score91.793.8right
Holdout score90.393.8right
Interval width8.59.2left
Cost per run$1.850$1.138right
Median latency2.4s2.1sright
Sample size168112left

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 source-span-cite \ --model claude-fable-5