Search papers, labs, and topics across Lattice.
FAIR at Meta, Inria, Sorbonne Universit茅, Institut universitaire de France, CERMICS 脡cole des Ponts ParisTech, PSL Research University
Meta AI (FAIR)1
0
2
Feedback Distillation boosts reasoning model performance by enhancing trajectory diversity and policy entropy, outperforming traditional methods like GRPO.
