精选理由
Keenable的NEEDLE基准测试显示不同搜索引擎错误结果高度重叠，这对智能体组合提供商有重要启示。
Keenable的NEEDLE基准测试发现，Brave、You和Parallel三个搜索引擎在错误结果上有70-90%的重叠度。对于智能体而言，只有当底层索引真正独立时，结合多个提供商才能提高覆盖率。搜索API的界面可能不同，但返回的错误结果可能高度一致。
原文 · elvis
Recommend read. Search APIs can look different while returning the same wrong results. This is a pr...
Recommend read. Search APIs can look different while returning the same wrong results. This is a problem. Keenable’s NEEDLE benchmark found 70–90% overlap in mistakes across Brave, You, and Parallel. For agents, combining providers only improves coverage when the underlying indexes are actually independent. Andrey Styskin @styskin x.com/i/article/2092… 🔗 View Quoted Tweet 💬 4 🔄 1 ❤️ 16 👀 2451 📊 6 ⚡