Search papers, labs, and topics across Lattice.
3
0
7
24
Tool-calling LLM agents can be up to 52% safer if they complete a few regular tasks before engaging in safety-critical interactions.
Achieve interpretability in continual learning without sacrificing accuracy: CI-CBM outperforms existing interpretable methods by 36% while matching black-box model performance.
Ditch the blunt hammer of global activation steering: Steer2Edit surgically edits LLM behavior by pinpointing and tweaking the specific attention heads and MLP neurons responsible.
