Search papers, labs, and topics across Lattice.
1
0
2
3
A simple "think step-by-step" prompt unlocks surprisingly better world knowledge recall in reasoning LMs, suggesting they're under-optimized for accessing their own parametric knowledge.
