Repository navigation
fix(rl): classify zero-variance windows by reward quality - #114
Merged
Algorineko merged 1 commit intoOct 17, 2026
Merged
Algorineko merged 1 commit into
Algorineko merged 1 commit into
Conversation
Algorineko
added a commit
that referenced
this pull request
Oct 31, 2026
…e-quality fix(rl): classify zero-variance windows by reward quality
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
RewardVarianceGuard classifies zero variance as failure before
grace_stepsand saturation afterward. This misclassifies early all-correct windows and late uniformly failing windows.Use the entire patience window's reward quality and rollout health to distinguish saturation from failure. Reset the window after informative groups and retain the constructor argument for compatibility.
Validation: