PulseAugur
中
实时 11:39:32

论文认为网络代理应采用编译执行而非ReAct

一篇新论文提出,网络代理应从主流的ReAct范式转向编译执行方法。作者认为,ReAct将运行时网络内容直接输入模型会增加被提示注入攻击的风险。相反,编译执行模型在观察网络内容之前会承诺执行一项特定任务的程序,从而将不受信任的数据隔离在预定义的执行图中。该论文分析了WebArena基准测试,发现大多数任务都与提议的范式兼容,并且有很大一部分可以在没有运行时LLM子程序的情况下执行。确定的主要障碍是缺乏网络交互的语义、类型化接口,目前这些接口依赖于通用的浏览器操作,如点击和滚动。 AI

影响 这项研究可以通过减轻提示注入风险来增强基于网络的AI代理的安全性和可靠性。

排序理由 学术论文,提出了一种新的AI代理范式。[lever_c_demoted from research: ic=1 ai=1.0]

在 arXiv cs.AI 阅读 →

AI 生成摘要 · Google Gemini · 来自 1 个来源。 我们如何撰写摘要 →

论文认为网络代理应采用编译执行而非ReAct

本文如何被排名

Signal score
10 / 100
Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket
Tool
学术论文,提出了一种新的AI代理范式。[lever_c_demoted from research: ic=1 ai=1.0]
Source corroboration
Single-source cluster
Only one publisher covered this so far. Single-source stories can still rank when the publisher is high-authority, but they lack cross-source corroboration.
Topics
paper, safety
Editorial topic classification. Feeds into how the story surfaces on /topic/<slug> hub pages and into the per-entity coverage mix.
AI-industry relevance
High
Clearly on-topic for AI-industry coverage.
Story freshness
Same-day
Cluster formed today. Ranking reflects the current source set at time of score.

完整方法见我们的编辑标准。

报道来源 [1]

  1. arXiv cs.AI TIER_1 English(EN) · Julien Piet, Annabella Chow, Yiwei Hou, Muxi Lyu, Sylvie Venuto, Jinhao Zhu, Raluca Ada Popa, David Wagner ·

    Web Agents 应采用编译执行范式

    arXiv:2605.14290v2 Announce Type: replace-cross Abstract: ReAct has become the default architecture across LLM agents, and many existing web agents follow this paradigm. We argue that it is the wrong default for web agents. Instead, web agents should default to compile-and-execut…