SMADE-IE: Sparse multi-agent debate framework for zero-shot IE

Arxiv CS.CL (NLP/language models)

Proposes SMADE-IE, a sparse multi-agent framework that uses evidence-driven debate among LLMs for zero-shot information extraction. Achieves state-of-the-art results on multiple benchmarks without task-specific training.

llmmulti-agent-systems

Arxiv CS.CL (NLP/language models)·Kenfeng HuangYi CaiXin WuZikun DengLi Yuan

7 days ago