Search papers, labs, and topics across Lattice.
Affiliation:
1
0
3
5
Restricting LLM fine-tuning to the bottom singular subspace nearly eliminates catastrophic forgetting, even when protecting 99% of weights, while maintaining strong target-task performance.