Search papers, labs, and topics across Lattice.
Xi'an Jiaotong Uni- versity
1
0
3
LLM agents can get a ~5% performance boost and 4x memory reduction by representing workflows as graphs and adaptively instantiating them based on task semantics.
