Search papers, labs, and topics across Lattice.
1
0
3
67
LLMs encode truthfulness along a spectrum from general principles to narrow domain expertise, and exploiting domain-specific truth representations is key to steering model behavior.
