Search papers, labs, and topics across Lattice.
2
0
5
3
LLMs can achieve a 7.5x performance boost in web search and extraction by using a bi-level multi-agent architecture with iterative refinement and shared memory.
Forget hand-crafting agents: Memento-Skills lets a generalist LLM agent autonomously design and improve specialized agents through experience, achieving substantial gains on complex benchmarks.
