Search papers, labs, and topics across Lattice.
1
0
2
LLMs can be made significantly more truthful by identifying and down-weighting the influence of hallucination-prone tokens during generation.
