Alex XU
Alex XU
Home
Skills
Experiences
Publications
Projects
Posts
Courses
Schedule
Light
Dark
Automatic
1
Evaluating a Terminal Task Is Itself a Terminal Task: What LLM Graders Need to Judge Agent Work
A poster-length version of Probe-and-Resume, studying what an LLM critic needs to grade terminal tasks when no verifier is available.
Tiancheng Ge
,
Alex Jiakai XU
,
Zhou Yu
,
Eugene Wu
,
Kostis Kaffes
PDF
Project
From Runnable Code to Shippable Applications: Test-Driven Development for Full-Stack Web Application Generation
Coding agents can generate runnable web applications, but their outputs frequently fail to satisfy functional requirements. Although …
Yuxuan Wan
,
Tingshuo Liang
,
Alex Jiakai XU
,
Jingyu Xiao
,
Yintong Huo
,
Michael R. Lyu
Cite
Project
DOI
arXiv
Toward Systems Foundations for Agentic Exploration
This paper presents a systems perspective on the challenges of supporting agentic exploration in LLM-powered agents, focusing on fork semantics, external side-effects, and native forking.
Alex Jiakai XU
,
Tianle Zhou
,
Eugene Wu
,
Kostis Kaffes
Cite
Project
DOI
Cite
×