3 papers indexed
Junming Chen, Junyang Jiang, Xu Chen, Zibo Liang, Kai Zheng
LLM-based database agents show promise, but differing task scopes, testbeds, and metrics hinder comparison. We identify four gaps between evaluation and production operations: live-environment fidelity (multi-turn read-write interaction with a running database); observation-space…
Wenjun Wang, Yuchen Fang, Fengrui Liu, Zibo Liang, Kai Zheng
Large Language Model (LLM) agents have shown remarkable capabilities in autonomous decision-making by generating sequential trajectories of states, actions, and observations. However, in complex, long-horizon tasks, these agents frequently suffer from compounding errors and strug…
Yuhang Zhou, Kai Zheng, Haoling Li, Dengyun Peng, Can Xu, Jingjing Chen
On-policy distillation (OPD) trains a student policy by matching a stronger teacher on the student's own trajectories, offering a promising framework for language agent training. However, its application to long-horizon agentic tasks remains insufficiently explored. We identify t…