Search papers, labs, and topics across Lattice.
2
0
5
6
Forget end-to-end fine-tuning: ChainFed unlocks private LLM adaptation on edge devices by training adapters layer-by-layer, boosting accuracy by up to 46%.
Unlearning in multimodal recommendation systems is broken because deleted data's influence isn't uniform; TRU fixes this with targeted interventions, achieving better forgetting with less performance loss.
