Search papers, labs, and topics across Lattice.
1
0
3
LLMs get schooled in dialogue state tracking by a mixture-of-experts architecture that uses a graph neural network and ReAct agents to achieve state-of-the-art results with a T5-Small backbone.
