Search papers, labs, and topics across Lattice.
1
0
3
RLHF might be sacrificing reasoning faithfulness in LLMs, even while boosting accuracy.