Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
1
0
3
Aligning language models with human preferences just got more stable: a new method avoids the divergence problems of direct density ratio optimization.