1
皮尤研究中心：ChatGPT 问世以来的新网页中，三分之一存在 AI 痕迹
皮尤研究：ChatGPT问世后新网页超三分之一含AI痕迹，互联网内容生态正在被重塑
IT之家 8 月 23 日消息，据外媒 TechCrunch 报道，当地时间 20 日，皮尤研究中心发布的一项研究显示，自 2022 年 11 月 ChatGPT 问世后发布的网页中，超过三分之一可能带有 AI 生成痕迹。这一结果与其他研究相互印证，表明 近年来新增的大量网页已经由 AI 直接生成，…
2
Claude Watermark
一键识别并清除文本中的AI痕迹，保护原创性与隐私的轻量工具
Find and remove every trace AI leaves in your text Discussion | Link
3
LLM Model-Swapping Trick Can Expose AI Reasoning Traces
模型替换攻击可撬开AI思维链，警惕推理过程泄露风险
Article URL: https://ai-updates.net/model-swapping-exposes-ai-reasoning-traces/ Comments URL: https://news.ycombinator.com/item?id=49265641 Points: 2 …
4
Stealing Reasoning Traces from Proprietary LLM APIs
一条加密token泄露链，暴露闭源大模型推理过程可被逆向窃取，安全边界再受拷问。
Stealing Reasoning Traces from Proprietary LLM APIs A vanity domain name ( stolen-thoughts.com ) for a neat paper : Anthropic, OpenAI, and Google retu…
5
Stealing Reasoning Traces from Proprietary LLM APIs
揭秘如何从专有大模型API中窃取隐藏推理过程，安全研究者必读。
Stealing Reasoning Traces from Proprietary LLM APIs A vanity domain name ( stolen-thoughts.com ) for a neat paper : Anthropic, OpenAI, and Google retu…
6
Stealing Reasoning Traces from Proprietary LLM APIs
揭秘API加密推理痕迹的可窃取性，安全研究必读。
arXiv:2608.09867v1 Announce Type: cross Abstract: Leading large language model providers now conceal their models' step-by-step reasoning, or chain-of…
7
Constraint-Anchored Reasoning Traces
15页论文提出约束锚定推理痕迹方法，为AI推理提供新视角，ACM MM 2026收录。
arXiv:2607.16727v1 Announce Type: new Abstract: Autoregressive multimodal large language models (MLLMs) suffer from error snowballing: a single incorr…
8
AI Engram: In Search of Memory Traces in Artificial Intelligence
在AI中寻找记忆印记，ICML 2026 Oral论文揭示大模型内部记忆痕迹机制。
arXiv:2606.14997v1 Announce Type: new Abstract: Memory formation is fundamental to intelligence, yet whether deep neural networks preserve identifiabl…
9
Counterfactual Trace Auditing of LLM Agent Skills
反事实追踪审计方法，可揭示LLM智能体技能的可解释性与可靠性缺陷，提升AI安全。
arXiv:2605.11946v2 Announce Type: replace Abstract: Large Language Model agents are increasingly augmented with agent skills. Current evaluation metho…
10
Diagnosing Harmful Continuation in Answer-Correct Long-CoT Training Traces
聚焦长链思维训练中“答案正确但有害”的隐蔽问题，提出诊断方法助你避开模型训练陷阱。
arXiv:2605.29288v1 Announce Type: new Abstract: Long chain-of-thought (CoT) traces are widely used as supervision for reasoning-oriented LLM SFT, yet …
11
ReasonOps: Operator Segmentation for LLM Reasoning Traces
新方法将LLM推理轨迹分割为操作符，提升可解释性与效率
arXiv:2605.29192v1 Announce Type: new Abstract: Chain-of-thought traces from large reasoning models can span tens of thousands of tokens, yet we lack …
12
Extracting Search Trees from LLM Reasoning Traces Reveals Myopic Planning
从LLM推理中提取搜索树，揭示其规划过程存在“近视”缺陷，对理解模型智能边界有重要启发。
arXiv:2605.06840v5 Announce Type: replace Abstract: Large language models (LLMs), especially reasoning models, generate extended chain-of-thought (CoT…
13
Confidence Geometry Reveals Trace-Level Correctness in Large Language Model Reasoning
置信度几何方法揭示大语言模型推理中痕迹级正确性的全新框架，附开源代码。
arXiv:2605.16824v1 Announce Type: new Abstract: Large language models (LLMs) generate not only reasoning text, but also token-level confidence traject…
14
Explainable Detection of Depression Status Shifts from User Digital Traces
用AI分析社交媒体时间序列，可解释地检测抑郁状态演变，为心理健康监测提供新方法。
arXiv:2605.14995v1 Announce Type: new Abstract: Every day, users generate digital traces (e.g., social media posts, chats, and online interactions) th…
15
Skyra: AI-Generated Video Detection via Grounded Artifact Reasoning
提出基于人工痕迹推理的AI视频检测方法，兼具高准确率与可解释性。
arXiv:2512.15693v2 Announce Type: replace Abstract: The misuse of AI-driven video generation technologies has raised serious social concerns, highligh…