Junjie Chu
Titles as the sources gave them. In 2 entries, with 0 checked quotations.
In the Base
Every entry they appear in.
Can Frontier Auto Researchers Reconstruct Human Research Decisions in AI?
According to the abstract, Shengyun Si and colleagues introduce AutoHRS, a stage-wise framework for measuring how closely AI auto researchers reproduce the decisions and outputs of human AI researchers. They generated 456 research trajectories from 152 ICLR oral papers; the excerpt gives no results.
GuardrailAgent: An Agentic Framework for Jailbreak Defense
A preprint proposes GuardrailAgent, a multi-agent defense that gathers attack characteristics into a continuously updated knowledge base to detect jailbreaks on input and output. The authors report a 93.1% input-side detection rate on over ten jailbreaks.