1
Clean up Claude 5's token vomit with a separate LLM
让本地小模型把Claude的“token呕吐”翻译成人话，全程本地无遥测，专治长篇AI语无伦次。
Article URL: https://github.com/zachahn/vomit Comments URL: https://news.ycombinator.com/item?id=49375996 Points: 60 # Comments: 50
2
张雪机车调整服务渠道，抖音私信不再处理用户咨询与售后
张雪机车将用户咨询与售后从抖音私信转移至官方小程序和App，提升服务效率与用户体验。
IT之家 7 月 16 日消息，张雪机车今日发布 服务渠道调整通知。为优化咨询及售后处理流程，提升订单、售后问题处理效率，团队对线上咨询服务渠道进行统一调整，相关重要变更通知如下：自 2026 年 7 月 17 日起， 抖音私信渠道 将不再回复任何用户咨询、订单查询、车辆售后、维修质保、故障反馈类问…
3
SAGE: Retain-Aware Post-Hoc Sanitization of Final Unlearning Vector
AI模型遗忘的安全隐患触目惊心，SAGE提出后处理方案在保留核心能力的同时彻底清除敏感数据
arXiv:2606.18309v1 Announce Type: new Abstract: Large Language Model (LLM) unlearning aims to remove undesirable knowledge or behaviors while preservi…
4
Iterative LLM-based improvement for French Clinical Interview Transcription and Speaker Diarization
多轮LLM迭代优化，法语医疗对话转录准确率与说话人归属大幅提升，突破30%词错误率瓶颈
arXiv:2603.00086v2 Announce Type: replace Abstract: Automatic speech recognition for French medical conversations remains challenging, with word error…
5
Fingerprinting LLMs via Prompt Injection
提出通过提示注入对LLM进行指纹识别，破解模型衍生关系判断难题，为AI安全溯源提供新思路。
arXiv:2509.25448v3 Announce Type: replace-cross Abstract: Large language models (LLMs) are often modified after release through post-processing such a…
6
Introducing the v0 composite model family
Vercel揭秘v0复合模型：融合RAG、LLM与后处理模型，显著提升代码生成质量
We recently launched our AI models v0-1.5-md , v0-1.5-lg , and v0-1.0-md in v0 . Today, we're sharing a deep dive into the composite model architectur…