Search papers, labs, and topics across Lattice.
1
0
2
Removing misleading reasoning from answer-correct CoT traces can drastically enhance fine-tuning performance in LLMs.