Search papers, labs, and topics across Lattice.
2
0
6
Adaptive layer selection in LLMs can significantly enhance inference efficiency without sacrificing accuracy, achieving better performance than static pruning methods.
LLMs can nail the final answer in code execution but still fail to reason about the steps to get there, exposing a critical flaw in current evaluation methods.
