20 days ago
-
Record v0.4.1 split materialization
-
Fix benchmark split materialization (v0.4.1)
-
Materialize benchmark_preferences_de test split (v0.4.1)
-
Materialize benchmark_preferences_de validation split (v0.4.1)
-
Materialize benchmark_preferences_de train split (v0.4.1)
-
Materialize benchmark_hard_negatives_de test split (v0.4.1)
-
Materialize benchmark_hard_negatives_de validation split (v0.4.1)
-
Materialize benchmark_hard_negatives_de train split (v0.4.1)
-
Materialize benchmark_repairs_de test split (v0.4.1)
-
Materialize benchmark_repairs_de validation split (v0.4.1)
-
Materialize benchmark_repairs_de train split (v0.4.1)
-
Record v0.4 benchmark leakage and test replay evidence
-
Release v0.4 family-disjoint benchmark layer
-
Add v0.4 test replay receipt
21 days ago
-
Add v0.4 benchmark manifest and leakage checks
-
Add v0.4 family-disjoint preference benchmark
-
Add v0.4 family-disjoint hard-negative benchmark
-
Add v0.4 family-disjoint repair benchmark
-
Add v0.4 family-disjoint hard-negative benchmark
-
Add v0.4 family-disjoint repair benchmark
-
Record v0.3 runtime and preference provenance
-
Release Evidence-Bound CSS v0.3 quality layer
-
Add v0.3 120 runtime-derived preference pairs
-
Add v0.3 120 verifier-confirmed hard negatives
-
Add v0.3 120 runtime-verified CSS repairs
-
Bump dataset card to v0.2.1 quality release
-
Record quality-layer provenance and browser admission gate
-
Document debug and browser-verified quality configs
-
Add first browser-verified CSS repair suite
-
Add 100 source-grounded CSS debug reasoning records
-
Restore full 601-record SFT config
-
Restore full 601-record knowledge config
-
Record publisher authorization and full corpus provenance
-
Release v0.2.0 full 601-record CSS corpus
-
Align provenance registry with live v0.1 seed
-
Align v0.1 card with published 32-recipe seed
-
Publish 32 curated German CSS solution recipes
-
Add source and validation registry
-
Add evidence-bound CSS v0.1 dataset card
22 days ago
-
initial commit
