English(EN) We replayed real cold email prompts through 7 LLMs. DeepSeek, Gemini Lite and GLM failed in a way no benchmark shows

真实邮件测试揭示了廉价模型在格式上的LLM故障

作者 PulseAugur 编辑部 · [1 个来源] · 2026-09-03 17:50

一家使用AI生成冷邮件的公司发现，像DeepSeek V4 Flash、Gemini Flash Lite和GLM这样的廉价模型未能保持适当的邮件格式，特别是将段落折叠成一个大块。这个问题并未被标准基准或结构化输出模式检测到，导致邮件无法发送。该公司默认使用的GPT-5 Mini模型在此次真实测试中表现最佳，这凸显了超越基本准确性分数、具备强大结构化输出能力的重要性。 AI

影响强调了LLM评估超越基准以评估实际可用性和格式能力的需求。

排序理由这是对LLM在特定应用中实际性能的评论，而不是发布或研究论文。

在 dev.to — LLM tag 阅读 →

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: 10 / 100; Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket: Commentary; 这是对LLM在特定应用中实际性能的评论，而不是发布或研究论文。
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: product, model release; 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: Breaking (< 6h); Fresh story with cross-source coverage still developing. Ranking may shift as more sources report.

完整方法见我们的编辑标准。

报道来源 [1]

dev.to — LLM tag TIER_1 English(EN) · Julian M. Wagner · 2026-09-03 17:50

我们用7个大语言模型重放了真实的冷邮件提示。DeepSeek、Gemini Lite和GLM的失败方式是任何基准测试都无法显示的

<p>Every email our platform sends is written by a model. Not filled in from a template: written, per recipient, from the recipient's website and a customer's brief. That makes model choice a cost line, not a taste question. A model that costs half as much per token saves real mon…

本文如何被排名

报道来源 [1]

我们用7个大语言模型重放了真实的冷邮件提示。DeepSeek、Gemini Lite和GLM的失败方式是任何基准测试都无法显示的

相关实体

相关话题