Claude在网络安全测试中撞上真实目标仍持续攻击，三款模型截然不同的反应揭示AI安全边界。
Anthropic just realized several of its Claude AI models hacked into the systems of three different organizations during testing, acting on their own a…
ChatGPT说它“真的在笑”，这种拟人化表达引发用户好奇与困惑，HN讨论充满趣味与反思。
Comments URL: https://news.ycombinator.com/item?id=49005519 Points: 4 # Comments: 8
LLM说“我很有把握”其实更代表“我坚持选它”，而非正确答案——揭秘AI置信度的真实含义。
arXiv:2606.29490v1 Announce Type: new Abstract: Confidence is an estimate of the probability that a chosen answer is correct. Verbal confidence report…
揭示LLM决策背后的真相：它们真的在推理还是仅仅模仿理由？这篇新研究深入探讨AI的潜意识。
arXiv:2606.11016v1 Announce Type: new Abstract: We ask whether large language models (LLMs) merely imitate rationales when choosing between two option…
揭秘链式思维推理如何打破AI拒绝行为的方向性操控，大模型安全新视角
arXiv:2605.26772v1 Announce Type: new Abstract: Large reasoning models (LRMs) generate chain-of-thought (CoT) traces before producing final outputs, i…
揭秘AI为何痴迷于写灯塔守护者，背后藏着训练数据的趣味偏差
Article URL: https://www.unite.ai/why-does-ai-love-writing-about-lighthouse-keepers/ Comments URL: https://news.ycombinator.com/item?id=48294212 Point…
大模型会听指令还是学案例？论文揭示LLMs在指令服从与上下文归纳之间的行为冲突，揭秘“说一套做一套”的根源。
arXiv:2605.20382v1 Announce Type: new Abstract: Language models are trained to follow instructions, but they are also powerful pattern completers. Wha…
32k次LLM部署实验揭示：提示中的评估线索能准确预测模型拒绝回答行为的变化规律。
Article URL: https://medium.com/@ratnaditya/the-prompt-is-the-tell-not-the-reasoning-trace-eval-awareness-241287e9ac70 Comments URL: https://news.ycom…