AI Agent 论文播报 6/15:Agent 运行时被严肃对待 episode artwork

EPISODE · Jun 16, 2026 · 9 MIN

AI Agent 论文播报 6/15:Agent 运行时被严肃对待

from 周六9点半

这一期我们聊一件正在发生的转变:Agent 的研究重心,正从「更聪明的模型」滑向「更可治理的系统」。今天的三篇重点论文,分别从安全、harness 演化和具身 scaffold 三个切口,把 Agent 运行时层推上了独立的研究台面。本期重点* 面向 Agent 浏览器的同源策略(Same-Origin Policy for Agentic Browsers):把 Web 经典的同源策略搬到 Agent 浏览器,指出 Agent 本身就是一条绕过 SOP 的跨域数据通道,并给出基准 SOPBench 与运行时防御 SOPGuard——用类似 OS 污点跟踪的思路,把跨域写入退化为「人工授权」一步。* HarnessX:可组合、可适应、可演化的 Agent 外壳铸造厂:把 prompt、工具、记忆、控制流这层 harness 当成一等对象,把 harness 改写形式化为 RL 问题,...去小宇宙查看完整单集简介在小宇宙查看该单集文稿

Episode metadata supplied by the publisher feed · Published Jun 16, 2026

Embed this episode

Ready to play

AI Agent 论文播报 6/15:Agent 运行时被严肃对待

0:00 9:34

No transcript for this episode yet

We transcribe on demand. Request one and we'll notify you when it's ready — usually under 10 minutes.

No similar episodes found.

No similar podcasts found.

Frequently Asked Questions

How long is this episode of 周六9点半?

This episode is 9 minutes long.

When was this 周六9点半 episode published?

This episode was published on June 16, 2026.

Can I download this 周六9点半 episode?

Yes. Use the download control on the episode player to save the publisher-provided media file.
URL copied to clipboard!