Search papers, labs, and topics across Lattice.
1
0
2
SelfCompact reveals that language models can autonomously manage context decay, achieving up to 18.1 points improvement in performance while cutting token costs by 30-70%.
