Search papers, labs, and topics across Lattice.
1
0
2
4
Task structure, not difficulty, dictates when reasoning in LLMs pays off, revealing surprising inefficiencies in common benchmarks.
