Search papers, labs, and topics across Lattice.
3
0
8
A unified assessment framework reveals hidden insights about agent performance, transforming how we evaluate AI systems.
Forget lab coats: MIND lets LLMs autonomously formulate and experimentally validate materials science hypotheses, potentially accelerating discovery.
Forget activation whitening – Hessian-aware decomposition unlocks 20-40% better low-rank LLM approximations.
