2 papers
cs.CR2026
Same Attack, Different Judge: A First Controlled Adversarial Comparison of Jev, LLM Judges, and Purpose-Built Safety Classifiers
Xiang Li, Aichun Huang, Mingming Zhang +1
TypeSafe AI's Jev is a new kind of safety judge. It generates no text and no label: it answers questions the developer defines in advance and returns a probability for each answer.…
cs.SE2026
When Talk Isn't Code: Comparing LLM Agents That Simulate and Develop Software
Xiang Li, Aichun Huang, Mingming Zhang +1
LLMs are used both to simulate network protocol implementations, as honeypots do, and to write them. Prior work evaluates the two uses separately, from a model's answers or convers…