1% of latency.
Lightweight decision model reduces token use in multi-agent systems by up to 54.6%
Jev-LiteMAS maintains accuracy within 1.34 percentage points of fully generative workflows while cutting latency in seven of eight evaluated settings.
Top University
Kaixiang Wang · Zhaijiacheng Zhou · Jiong Lou · Jie Li
Shanghai Jiao Tong University
Research Digest··3 min read
Wang et al.
§