跳到正文
Fintech News Singapore· Fintech News Singapore·· 16天前AI 评分48

Visa:AI模型近半数测试未能作出合适购物决策

Visa: AI Models Failed Nearly Half of Tests to Make Suitable Shopping Decisions

AI 导读

AI-assisted source summary · not independently verified. Content standards

Visa的Agentic Commerce Benchmark显示,22个AI模型在日常购物测试中近半数尝试失败,整体通过率为57.2%。测试覆盖12家AI实验室的模型和15个涉及电视、笔记本电脑及个人护理电器的购物场景;最佳模型通过率82.2%,最低为28.9%。此外,54%的失败尝试仍促成购买,Visa计划纳入结账、支付和下单测试。

Source:Fintech News Singapore · fintechnews.sg