Search papers, labs, and topics across Lattice.
1
0
3
Model merging can be made significantly more robust to distribution shift by explicitly modeling uncertainty and using contrastive learning to align consistent samples and separate conflicting ones.
