技能是双刃剑：研究揭示LLM代理的技能提升可能伴随任务回归代价，近6000次实验拆解正负效应。
arXiv:2607.22520v1 Announce Type: new Abstract: Adding procedural skills to an LLM agent is typically evaluated by average improvement in task success…
AI让一句话变成五页备忘录？警惕认知层面的"大规模杀伤性武器"。
Article URL: https://taoofmac.com/space/blog/2026/07/08/2130 Comments URL: https://news.ycombinator.com/item?id=48842713 Points: 3 # Comments: 0
从副作用的根源剖析LLM行为干预，低秩子空间分析精准定位，为安全控制提供新视角
arXiv:2606.14388v1 Announce Type: new Abstract: Interventions designed to modify a particular behavior in LLMs, such as refusal or sycophancy, often p…
普华永道报告揭示：AI病历记录工具因精准记录更多诊断细节，正成为推高医疗费用的隐形推手，2027年医疗成本或增9%。
Article URL: https://fortune.com/2026/06/12/ai-making-medical-bills-higher/ Comments URL: https://news.ycombinator.com/item?id=48519519 Points: 33 # C…
响应式编程为何从热门变遗留? 一位开发者亲述维护RxJS项目的困境与思考。
I've recently been put on a legacy project that uses reactive programming and RxJS. This is the first time I'm coming across this programming paradigm…