English(EN) Each time I test Gemini with a specific question, it’s half, fully, or catastrophically wrong. Only *once* was it fully correct, for a French BBQ sauce—but when

用户报告Gemini频繁出错，捏造细节

作者 PulseAugur 编辑部 · [1 个来源] · 2026-05-17 20:47

一位Mastodon用户分享了他们测试Google的Gemini模型的负面经历。他们报告说，当被问及具体问题时，Gemini经常出错，有时甚至是灾难性的错误。在极少数它正确的情况下，重复相同的提示但带有轻微的拼写错误会导致捏造信息，这凸显了其可靠性方面的不足。 AI

影响用户报告表明Gemini的响应可能不可靠，影响了信任和采用。

排序理由用户对模型性能的意见，而非可验证的基准或官方发布。

在 Mastodon — fosstodon.org 阅读 →

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

报道来源 [1]

Mastodon — fosstodon.org TIER_1 English(EN) · [email protected] · 2026-05-17 20:47

Each time I test Gemini with a specific question, it’s half, fully, or catastrophically wrong. Only *once* was it fully correct, for a French BBQ sauce—but when

Each time I test Gemini with a specific question, it’s half, fully, or catastrophically wrong. Only *once* was it fully correct, for a French BBQ sauce—but when I asked again but mistyped 1954 as 1965, it produced an equally elaborate marketing story about a BBQ sauce that doesn’…

报道来源 [1]

Each time I test Gemini with a specific question, it’s half, fully, or catastrophically wrong. Only *once* was it fully correct, for a French BBQ sauce—but when

相关实体

相关话题

Each time I test Gemini with a specific question, it’s half, fully, or catastrophically wrong. Only once was it fully correct, for a French BBQ sauce—but when