Search papers, labs, and topics across Lattice.
Zhejiang University
1
0
3
DeferMem's RL-trained memory distiller lets LLMs cherry-pick and rewrite scattered information into query-specific evidence, slashing memory costs and boosting QA accuracy.
