Repository navigation
fix: honor zero thresholds when realizing agent graphs - #39
Closed
liuzhengyang699 wants to merge 1 commit into
Closed
liuzhengyang699 wants to merge 1 commit into
liuzhengyang699 wants to merge 1 commit into
Conversation
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Passing
threshold=0.0toEdgeWiseDistribution.realizecurrently takes the stochastic branch because the threshold is checked for truthiness. A positive-probability edge can therefore disappear even though its probability is above the requested threshold. Learned node ordering already treats zero as an explicit threshold, making the two decisions inconsistent.Check for
Noneexplicitly so zero uses the same thresholding logic as other values.Validation: 6 focused offline tests cover thresholds zero, one, and
None, with and without learned ordering. The two zero-threshold cases fail before the fix. Tests are markedmock_llmand were run withpython -m pytest -m mock_llm test/swarm/optimizer/test_edge_threshold.py -q.git diff --checkpassed.