Xingzhou Chen

Behavioral Controllability of Agentic Models for Information Extraction: From Fixed Workflows to Reflective Agents

Lujia Zhang, Xingzhou Chen, Hongwei Feng

Large language model (LLM) agents are increasingly used for complex information-extraction tasks, yet it remains unclear whether agentic components such as reflection and memory lead to observable and controllable improvements over fixed LLM workflows. We study this question thro…Read more