Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
2
7
5
14
Exploiting local parameter perturbations of a teacher model unlocks complementary supervision signals, boosting LLM mathematical reasoning accuracy beyond standard on-policy self-distillation.
By recognizing that not all tokens are created equal, D2PO offers a simple temporal weighting fix that boosts DPO alignment scores by up to 9.7 points.