test(eval): sharpen java held-out fixtures and calibrate baseline (#68)#73
Merged
Conversation
The held-out java fixtures were agent-authored in #15 and never real-model validated; they were ambiguous, so a real model scored the split 0.25/0.25 and tripped the seen-vs-held-out overfit FLAG — a fixture-quality problem, not a profile-overfit one. Rewrite them as unambiguous, realized defects, each with a clean precision twin: - BugPathTraversal: realized Files.readAllBytes(...) of the user-controlled path (was a bare new File(...) constructor a model reasonably won't flag) -> cat#3 - BugSwallowedInterrupt -> BugSwallowedParse: swallowed NumberFormatException to a sentinel, no concurrency flavor, so no cat#2-vs-cat#5 ambiguity -> cat#2 - CleanSafePath: path read behind a strict whitelist (precision twin for cat#3) - CleanReportedParse: parse error propagates, nothing swallowed (twin for cat#2) Recalibrate baseline.held-out.txt to the measured 1.00/1.00 (runs=3, ±0.00, clean_fp_runs=0, gate PASS) now that the runners are deterministic (#71/#72). The seen-vs-held-out gap closes, so the overfit FLAG correctly clears. Closes #68 Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
This was referenced Jun 17, 2026
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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
Sharpen the java held-out eval fixtures and calibrate the split's baseline.
The held-out fixtures were agent-authored in #15 and never real-model validated. They
were ambiguous, so a real model scored the split
0.25/0.25and tripped the seen-vs-held-outoverfit FLAG — a fixture-quality problem, not a profile-overfit one (the same runner
fix made rust/shell/yaml/swift perfect).
Changes
Every defect is now realized and unambiguous, and each buggy fixture has a clean
precision twin (mirrors the seen split's 1.00-scoring balance):
BugPathTraversalFiles.readAllBytes(Paths.get("/data/" + name))— an actual read, not a barenew File(...)constructorBugSwallowedParse(wasBugSwallowedInterrupt)NumberFormatException→sentinel; no concurrency flavor, so no cat#2-vs-cat#5 splitCleanSafePath(wasCleanRestoresInterrupt)CleanReportedParse(new)Calibration
baseline.held-out.txtrecalibrated from the conservative placeholder (0.80/0.50) to themeasured
1.00/1.00(runs=3,±0.00,clean_fp_runs=0, gate PASS), now that therunners are deterministic (#71 / #72). Floors and bands unchanged. The seen-vs-held-out gap
closes → the overfit FLAG correctly clears.
Depends on
#72 (runner determinism) — already merged to
dev. Without it the1.00/1.00baselinewould not be reproducible.
Closes #68
🤖 Generated with Claude Code