Search papers, labs, and topics across Lattice.
Affiliation:
2
0
4
LLMs remember too much to be good user simulators, but targeted prompting and a novel "compactor" can make them forget like humans do.
