If I read your results correct it solved 5% more problems and used up between 3-5% fewer tokens, with a faster response time?
Correct. But the 5% more solves and 3-5% fewer tokens are within the noise of previous eval runs I've done on models. The faster response time is not.
Someone using an actual statistical test to quantify if there's actual measurable improvement?! An angel among men.
from my basic stats knowledge, it's not reliable to calculate the confidence interval of the sample proportion with fewer than about 10 samples on both sides. So if you're only seeing 1 or 2 failures in 20, and that approximately held as you scaled, you would have to run about 200 tests with each in order to get at least 10 failures in order to calculate the confidence intervals of the two samples and see if they overlap or not.
Re:Claude Fable 5.2 and Mythos 5.2
by disgruntledphd2 on Monday November 09, 2026 @04:03PM
(#50432094)
And this is why not many people do this. Even if you have the decals and the statistical knowledge, it's pretty expensive to do this regularly.
Uncertainty from count statistics are far larger than this. These are two perfectly consistent results there’s no statsig difference
The spending limit definitely goes down much faster as of last week or so.
Personally, I've found myself getting lazier recently, trying to get a single session to handle tasks that should be broken into multiple, etc. I've suspected claims of nerfing might be from changes in user habit. I'm glad more disciplined checks are being run.
Eval context: 30 bug fixes and features evaluated by tests the agent didn't see. Ran on Opus 5.5 at medium effort in Claude Code.
20 of the tasks were last run on Nov 1, twice each. Nov 1 solved 36/40, and today solved 19/20. Thinking tokens per task were 0.95x Nov 1’s and output tokens 0.97x. The one significant change was speed: median time to first token decreased from 3.0s to 1.7s.
Caveats: one run per task today, medium effort only.