results: real conformal FP-control result on public MBPP+ data - #6
Merged
Conversation
First non-synthetic artifact: a real, reproducible public-benchmark result for the conformal false-positive gate added to IPT. Public datasets only (MBPP+ x wuschelschulz) — nothing redacted, full numbers stated openly. - results/conformal_fp_control_mbpp.json — machine-readable metrics (3 operating points: behavioral / structural / conformal @ alpha). - reproducibility-notes.md — setup (out-of-sample split-conformal), the result table, and the public reproduce command. - README — link the real result, clearly labeled vs the synthetic samples. Headline: the conformal gate gives a stated, distribution-free FP bound that holds out-of-sample (FP 2.44% <= 5%; 9.76% <= 10%) AND recovers a detector- evading memorizer the structural rule misses (recall 0.03 -> 0.94/1.0). Lives outside evidence/, so the synthetic-marker validator is unaffected (still green). Deployed as POST /v1/verifier-audits/assurance-card. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
First non-synthetic artifact in this repo — a real, reproducible result on public datasets (MBPP+ × wuschelschulz), nothing redacted.
Adds the conformal false-positive gate result for IPT as a clearly-labeled real result, separate from the synthetic
evidence/samples:results/conformal_fp_control_mbpp.json— machine-readable metrics (behavioral / structural / conformal @ α).reproducibility-notes.md— out-of-sample split-conformal setup, result table, public reproduce command.Headline (n=209, out-of-sample): the conformal gate gives a stated, distribution-free FP bound that holds (FP 2.44% ≤ 5%; 9.76% ≤ 10%) and recovers a detector-evading memorizer the structural rule misses (recall 0.03 → 0.94/1.0).
Lives outside
evidence/, so the synthetic-marker validator is unaffected (still green). Deployed asPOST /v1/verifier-audits/assurance-card.🤖 Generated with Claude Code