Repository navigation
fix: prove native conditioning before preserving reasoning markers - #974
Draft
bradhilton wants to merge 1 commit into
Draft
bradhilton wants to merge 1 commit into
bradhilton wants to merge 1 commit into
Conversation
bradhilton
changed the title
This branch has not been deployed
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
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Agent: Hayek
A template-specific exception cannot establish whether literal reasoning markers belong to sampled assistant output. Replace it with a behavior check: render copies and verify the full native prefix before preserving that content with thinking explicitly disabled.
Missing evidence or unsupported rendering declines the repair. This does not reinterpret SFT or reasoning data, and sampled stop-token handling in #973 is separate.
Validation: 536 CPU checks and six public Qwen BPE cases passed, along with retained CI. The proof has measurable cost: ordinary thinking-off cases add 12–37.5%, and a 32-turn all-literal case takes 2.52 times as long. This is a correctness change, not a performance improvement.