Search papers, labs, and topics across Lattice.
Netflix
1
0
2
Mult-DPO reveals that a multinomial approach can effectively align LLMs with complex user preferences, outperforming traditional pairwise methods.
