Marius CiepluchandClaude Fable 5.1 3d08d692ff bench(train): class-targeted sourcing, --family, from the public OSV dumps ...

sources --family <family> keeps only advisories whose CWE the taxonomy maps to that family, writes
sources_<family>.json beside the training set, and excludes the pilot, every repository the
benchmarks already name, the strict used set and every earlier batch; --prefilter-only runs the
dump pass without GitHub. Implemented by Codex (gpt-6-astra), gated here: names normalize rejects
exclude nothing instead of aborting, and the strict used set is in the exclusions.

Co-Authored-By: Claude Fable 5.1 <[email protected]>

2026-10-13 21:46:45 +02:00

..

archive

ci: green again -- format and lint the whole tree; push the scan and runner images to GHCR on tags

2026-10-02 20:09:28 +02:00

ax

search-with-proof: no-model demo replay (executor + verifier on a stored demo)

2026-10-06 14:50:14 +02:00

dev

ci: green again -- format and lint the whole tree; push the scan and runner images to GHCR on tags

2026-10-02 20:09:28 +02:00

experiments

learn: exp-005 graph slice -- inconclusive (within-pair AUC 0.762 -> 0.746, CI [-0.063, 0.032]); unsure rate 0.29 -> 0.24; fill \$1.59

2026-10-03 16:58:37 +02:00

fixtures

feat(ruleset): PHP quick-mode rules from the engine's vocabulary, measured on the development corpus

2026-09-30 15:05:47 +02:00

independent

refactor(p1): one source-facts module owns function ranges, callers, declaration tables and language aliases

2026-10-13 12:00:37 +02:00

learn

refactor(p1): one source-facts module owns function ranges, callers, declaration tables and language aliases

2026-10-13 12:00:37 +02:00

manifests

feat(ruleset): PHP quick-mode rules from the engine's vocabulary, measured on the development corpus

2026-09-30 15:05:47 +02:00

measurements

measure: final push-hook arm on all 252 pilot changes (0/12, 2.1 false alarms per 100) and the combined whole-tool score

2026-10-13 15:50:29 +02:00

pairs

fix(p1,p0): restore the pre-P1 range contract for two callers; docs link and maintainer paths

2026-10-13 16:15:32 +02:00

push

chore(p0): drop never-enabled hunter tools, test-only functions and receiver delivery

2026-10-13 11:06:09 +02:00

ranking

feat(pre-push-safety-net): preserve uncertainty in benchmark scoring

2026-09-12 19:04:20 +02:00

repos

feat(pre-push-safety-net): pin labeled evaluation inputs

2026-09-12 18:45:54 +02:00

search

spec+docs(search-with-proof): record first cluster demonstration + fixed-side negative [skip ci]

2026-10-07 13:25:52 +02:00

unseen

bench(valid): a training freeze may cap the history walk at the newest N first-parent commits

2026-10-13 20:47:23 +02:00

valid

bench(train): class-targeted sourcing, --family, from the public OSV dumps

2026-10-13 21:46:45 +02:00