AI 大模型

AI Agent

智能体、工具调用、规划、工作流、多智能体和自主任务执行。

2026-02-23 至 2026-02-23 共收录 3 篇信号源：cs.AI, cs.CL, cs.LG, cs.SE

1. GUI与网页智能体 3 篇

2602.03175 2026-02-23 cs.LG 57%

Probe-then-Commit Multi-Objective Bandits: Theoretical Benefits of Limited Multi-Arm Feedback

探查-然后承诺多目标老虎机：有限多臂反馈的理论优势

Ming Shi

机构 * Department of Electrical Engineering, University at Buffalo（电气工程系，布法罗大学）

专题命中 GUI与网页智能体：agent(abstract)；分类 cs.LG

AI总结本文提出PtC-P-UCB算法，通过有限探查提升多目标老虎机的性能，实现$1/\sqrt{q}$的加速效果。

2507.08831 2026-02-23 cs.CV cs.LG cs.RO 57%

View Invariant Learning for Vision-Language Navigation in Continuous Environments

连续环境中视觉语言导航的视角不变学习

Josh Qixuan Sun, Huaiyuan Weng, Xiaoying Xing, Chul Min Yeum, Mark Crowley

机构 * Department of Electrical and Computer Engineering, University of Waterloo（电气与计算机工程系，滑铁卢大学）； Department of Civil and Environmental Engineering, University of Waterloo（土木与环境工程系，滑铁卢大学）； Department of Electrical and Computer Engineering, Northwestern University（电气与计算机工程系，西北大学）

专题命中 GUI与网页智能体：agent(abstract)；分类 cs.LG

AI总结本文提出VIL框架，通过对比学习和教师-学生机制提升连续环境中视觉语言导航的视角鲁棒性，实验表明其在多个基准数据集上性能优越。

Comments This paper is accepted to RA-L 2026

2602.18424 2026-02-23 cs.CV cs.RO 50%

CapNav: Benchmarking Vision Language Models on Capability-conditioned Indoor Navigation

CapNav：基于能力条件的室内导航基准测试

Xia Su, Ruiqi Chen, Benlin Liu, Jingwei Ma, Zonglin Di, Ranjay Krishna, Jon Froehlich

专题命中 GUI与网页智能体：agent(abstract)

AI总结 CapNav基准测试评估VLMs在不同移动约束下室内导航能力，发现其性能随约束增加而下降，且先进模型仍难以处理空间推理任务。