30 days ago
-
feat(h1b): add span factorial arms and offline probability replay tools
-
fix(mp-opd): exclude unalignable responses with reason metrics
-
feat(kd): add TRUST R B P and separate SimCT paper scores
-
feat(align): port isolated ALIGN-NEW with donor gradient parity
-
feat(simct): add isolated author baseline with loss gradient parity
-
fix(mp-opd): preserve GRASS noise state during evaluation
-
feat(mp-opd): port GRASS chunk ALIGN and partition baselines
-
chore(upstream): import Cross-Tokenizer-OPD source snapshot