registrycompare

XML fence guard against Role then format

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

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

xml-fence-guard

v2 by @dara
88.7
83.0 to 92.6 · holdout 87.5

role-then-format

v2 by @ilya
91.7
86.5 to 95.0 · holdout 90.3
Intervals overlap. The 3.0 point gap is not significant at this sample size.
Metricxml-fence-guardrole-then-formatBetter
Public score88.791.7right
Holdout score87.590.3right
Interval width9.68.5right
Cost per run$1.850$1.850same
Median latency2.0s2.4sleft
Sample size168168same

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 \ xml-fence-guard role-then-format \ --model claude-fable-5