Search papers, labs, and topics across Lattice.
Affiliation:
3
0
5
Constant learning rate combined with weight decay is mathematically incapable of maintaining a stable interior equilibrium in normalized networks, driving recurrent instabilities that can be precisely mapped and controlled via a single scalar law.
LLMs aren't just cheaper annotators; they can actually be *better* than humans at predicting aggregate opinions in certain subjective tasks.
Stop choosing between AI alignment and performance: this ensemble method dynamically switches between aligned and complementary models to maximize human-AI team performance.
