English(EN) My LLM Critic Disagreed With Itself on Every Trial. The Safe Part Was the Code I Didn’t Trust It to Touch.

大语言模型代理的批评者表现出极端的非确定性，但基于代码的安全措施确保了稳定性

作者 PulseAugur 编辑部 · [1 个来源] · 2026-08-28 07:15

一个大语言模型代理的内部批评者表现出极端的非确定性，在多次试验中对相同的输入产生不同的输出。这种不稳定性令人担忧，因为代理的安全性依赖于其一致性。然而，系统的整体安全性得以维持，因为关键的安全功能已从大语言模型批评者移至确定性代码，例如结构性低报检查和严重性分类允许列表。 AI

影响强调了大语言模型代理中基于确定性代码的安全措施的关键需求，因为仅靠模型的连贯性是不够的。

排序理由该项目讨论了大语言模型代理的稳定性和安全机制的研究，包括基准测试结果和架构更改。[lever_c_demoted from research: ic=1 ai=1.0]

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: 46 / 100; Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket: Tool; 该项目讨论了大语言模型代理的稳定性和安全机制的研究，包括基准测试结果和架构更改。[lever_c_demoted from research: ic=1 ai=1.0]
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: safety, product; 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。

dev.to — LLM tag TIER_1 English(EN) · Debashish Ghosal · 2026-08-28 07:15

我的大语言模型评论员在每次测试中都自相矛盾。安全的部分是那些我不信任它触碰的代码。

<blockquote> <p>In the <a href="https://dev.to/debashish_ghosal/my-agent-refused-96-times-that-was-the-right-output-1mg">My Agent Refused 96 Times. That Was the Right Output.</a>, I argued that the most valuable output from an agent planner is often a well-structured refusal. Thi…