Search papers, labs, and topics across Lattice.
Department of Intelligence and Information, Seoul National University
1
2
2
2
Current function-calling benchmarks are too simple: DICE-BENCH reveals that LLMs still fail at realistic, multi-turn tool use.
