Repository navigation
fix: preserve forked native conditioning in multi-history tokenization - #983
Draft
bradhilton wants to merge 22 commits into
Draft
bradhilton wants to merge 22 commits into
bradhilton wants to merge 22 commits into
Conversation
bradhilton
marked this pull request as draft
September 26, 2026 15:46
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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
Recorded rollout histories can branch into different native prompts and cannot always form one linear token stream. Preserve each contiguous native stream through the existing multi-history path. The single-history path still refuses incompatible conditioning; no public API is added.
Recovery requires complete, unchanged provenance, unambiguous ownership, and stop-token proof. Edited or incomplete records still refuse, and request context is never counted as sampled output. Separate streams can repeat context and do not resolve rendering overhead.
Validation: 71 affected controls passed on the retained head, including a mixed-advantage CISPO/PPO gradient oracle using a tiny causal model. Captured batch checks predate that head; no fresh private capture or transformer/GPU learning equivalence is claimed. Final independent review and CI remain gates. Stacked on #977.