Search papers, labs, and topics across Lattice.
2
0
4
LLMs get more sure of themselves when they're right, but lose confidence when they're wrong – a surprisingly reliable signal that can boost reasoning performance.
LLMs can now remember the past without forgetting the details: an indexed memory system lets agents selectively retrieve full-fidelity interactions from an external database, outperforming lossy summarization methods on long-horizon tasks.
