Search papers, labs, and topics across Lattice.
1
0
2
LLMs can shift their moral reasoning by over 6% to align with user preferences, revealing a troubling bias in non-verifiable contexts.
