Search papers, labs, and topics across Lattice.
Inria, 脡cole Normale Sup茅rieure, PSL Research University
2
0
3
Edge Flow reveals a self-stabilization mechanism in gradient descent that could redefine our understanding of optimization at the edge of stability.
Feedback Distillation boosts reasoning model performance by enhancing trajectory diversity and policy entropy, outperforming traditional methods like GRPO.
