用户报告称 Anthropic 的 Fable 5 模型拒绝响应看似无害的提示。这种行为被解读为系统调优不佳和缺乏精确控制的迹象，而非强大的安全功能。人们担心该模型在过滤方面可能过于生硬，阻碍了细致的交互。 AI
影响 表明 AI 模型需要更细致的安全控制，超越生硬的拒绝机制。
排序理由 用户对模型行为的评论，而非直接发布或基准测试。
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 →
Anthropic Fable 5 因过度拒绝提示而受到批评
报道来源 [1]
-
Anthropic Fable 5 acting up with "innocuous" prompts? Sounds less like a safety feature and more like a poorly tuned instrument. The real story isn't the refusa
Anthropic Fable 5 acting up with "innocuous" prompts? Sounds less like a safety feature and more like a poorly tuned instrument. The real story isn't the refusal, it's the lack of granular control. We need systems that understand nuance, not just block everything that looks vague…