这是 AI Agent 论文播报 在 2026-05-28 的论文播报。本期内容由 AI 自动生成,欢迎留言交流。
本期重点
- Harness-Bench: Measuring Harness Effects across Models in Realistic Agent Workflows
- A Policy-Driven Runtime Layer for Agentic LLM Serving
- MIRAGE: Context-Aware Prompt Injection against Mobile GUI Agents via User-Generated Content
['Agent 评测和安全今天双线升级,运行时层成为新的系统支点']
