docs(validator): record context limits and bounded fidelity study - #119
Merged
Merged
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
What and why
Publish the sanitized 54-request Qwen/Mac result without calling wrong-context comparisons detections. Record missing coverage, oversized events and successful offline evasion fixtures. Add a fixed September 17-24 follow-up window with context gates, held-out reporting and bounded calls; keep customer reliability and compensation separate from research authority.
Verification
Re-ran the retained independent verifier read-only: native bindings, source/capture hashes and metric reconstruction pass. Existing exclusive-create evidence was preserved. Relative Markdown targets and git diff checks pass. Staged and full HEAD-history gitleaks scans pass (155 commits). No new inference calls or runtime changes.
Risk and limits
Documentation only. Private prompts, raw captures, endpoints and operator identities are not published. Same-Mac repetitions are not independent references. The cheating rows are offline metric fixtures, not full-protocol live exploits. No model-identity, population-accuracy or economic-authority claim.