Search papers, labs, and topics across Lattice.
1
0
3
Stop throwing away information: Listwise preference optimization extracts more signal from multi-candidate image datasets and continuous reward scores, boosting diffusion model alignment beyond pairwise methods.
