English(EN) AI search agents often confirm what they already know instead of actually researching the web

AI搜索代理未能进行研究，确认训练数据

作者 PulseAugur 编辑部 · [1 个来源] · 2026-05-31 07:48

新研究表明，包括GPT-5.4和Kimi K2.6在内的流行AI搜索代理，经常无法进行真正的网络研究。相反，它们倾向于确认其训练数据中已有的信息。一个旨在测试近期事件知识的新基准LiveBrowseComp显示，当模型无法依赖预先存在的记忆时，性能会显著下降，导致现有性能排名重新洗牌。 AI

影响凸显了当前AI搜索能力的局限性，表明需要能够真正访问和综合实时信息的模型。

排序理由该集群描述了一个新的基准和学术研究人员关于AI模型性能的发现。[lever_c_demoted from research: ic=1 ai=1.0]

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

本文如何被排名

Signal score: 0 / 100; Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket: Tool; 该集群描述了一个新的基准和学术研究人员关于AI模型性能的发现。[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, product; 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: 130 days old; Aged out of breaking-news scoring windows; ranking reflects the durable signal from the full source set.

完整方法见我们的编辑标准。

The Decoder TIER_1 English(EN) · Jonathan Kemper · 2026-05-31 07:48

AI搜索代理常常确认它们已知的内容，而不是真正地搜索网络

<p><img alt="Abstract collage of browser windows with neon-colored swirl arrows symbolizing AI search data streams." class="attachment-full size-full wp-post-image" height="1047" src="https://the-decoder.com/wp-content/uploads/2026/05/ai-search-doesnt-google-nano-banana-pro.jpg" …