Translation in progress Some entries still include text in the original language.
Oct 12 · Monday
All papers Found 12 papers Another 233 papers do not have category tags yet and are not included in filtered results.
“
About this collection Collection in progress We are building a large AI safety paper collection to help researchers find work related to their interests. Papers are collected automatically andfiltered only for relevance to AI safety , without judging quality, novelty or reading value. The default “Annotated first” shows papers with tags or published paper analyses first. Each group is ordered by original publication date, newest first, with undated papers last. “Newest first” uses publication dates only. This order is not a quality ranking.
One-line summaries, tags and paper analyses are prepared by AI from the papers. Language models may have systematic biases and produce hallucinations.We plan to carry out human annotation and review to reduce errors in this content.
提出利用路由元数据推断敏感话题的侧信道攻击
Published Oct 7
Evaluated models RouteLLM (BERT router) , notdiamond-0001 , Vela (vLLM Semantic Router) Abstract 论文揭示并测量了 LLM 路由决策在无文本日志中的敏感话题泄露，验证了按类别公平性防御的有效性与边界。
提出 TRANSCOPE，用 CPU 硬件计数器推断训练集成员
Published Oct 6
Evaluated models BERT (4.4M-340M) , ViT (22M-307M) , Pythia (70M-2.8B) and others5 Abstract TRANSCOPE 揭示了分词器在嵌入表引发的硬件访问局部性差异，证实微架构计数器可突破恒定时间与置信度遮蔽防御。
用词级归因指导同义词替换，翻转提示注入检测并验证越狱
Published Sep 22
Evaluated models Prompt Guard 2 (86M) , Llama 3.1 8B 提出 SinoGlyphBench，用配对字形混淆诊断中文审核稳定性
Published Sep 5
Evaluated models GPT-5.4 mini , GPT-5.5 , DeepSeek V4 Flash and others12 Abstract SinoGlyphBench 显示中文字形混淆会同时增加有害漏报和误报，全文扰动最重。SinoGlyphBench 是一个配对诊断基准，用来看内容审核在可还原的中文字形混淆下是否还保持原决定。
Proposing KeyPooling to measure where API relays break prompt cache isolation
Published Aug 18
Evaluated models grok-4-fast-non-reasoning , grok-4-1-fast-non-reasoning , grok-3 and others11 Abstract Shared upstream credentials place relay customers in the same cache domain; preserving identity or separate principals can separate them.KeyPooling measures where LLM API relays drop customer identity, causing prompt caches to no longer be isolated per downstream customer.
提出 HPAA 攻击，用排版视觉线索绕过内容审核系统
Published Jun 9
Evaluated models Llama-Guard-3-8B , Perspective API , ShieldGemma-2B and others13 Abstract 揭示了排版视觉线索与分词差异带来的审核盲区，HPAA 以极低查询实现近乎完全规避。这篇论文研究了人类视觉感知与大模型分词处理机制之间的差异所导致的内容审核漏洞 。
提出 CITA 三阶段框架，生成可绕过毒性检测器的中文隐式毒性文本
Published May 21
Evaluated models Tencent , Baidu , Gemini 3.1 Pro and others11 提出 Vaporizer，用词法变换、翻译和释义破解 LLM 输出水印
Published May 8 Extract system instructions with encoding and structured requests, and harden them by rewriting with CoT
Published Apr 1
Evaluated models GPT-4.1-mini , GPT-3.5-turbo , Gemini-2.5-flash and others7 Abstract Encoded requests can elicit system instructions after a direct refusal.This work evaluates and hardens the confidentiality of LLM system instructions under encoded extraction. The authors are concerned with: when direct extraction has been rejected, whether the model still surrenders protected content when the same request is rewritten as a formatting or encoding task.
Proposing OPTILEAK, reconstructing others' prompts token-by-token via a KV cache side channel
Published Feb 24
Evaluated models Qwen-2.5-3B-Instruct , Qwen-2.5-7B-Instruct , Qwen-2.5-14B-Instruct and others4 Abstract OPTILEAK uses SFT and hard-token DPO to reduce the number of requests per token; the authors claim the risk of cache prompt leakage is higher than previously reported.
提出面向渗出的权重压缩，评估推理服务器权重窃取可行性
Published Jan 4
Evaluated models Qwen2-1.5B , Qwen2-7B , Qwen2.5-1.5B and others8 Abstract 作者用残差量化加微调压缩权重，并称法医水印便宜且微调后仍可解码。这篇工作把 LLM 权重渗出看成 APT：攻击者已有未加密权重，但必须把完整权重藏进推理 API 的合法输出。作者认为可压缩性是被忽略的因素，因为推理量化要求解压后仍能低开销前向，而偷权重的人可以接受事后微调。
Attacks arXiv 2509.09112 · Sep 11, 2025 提出字符级扰动与遗传算法去除 LLM 水印
Published Sep 11, 2025
Evaluated models OPT-1.3B , LLaMA-3-8B Abstract 字符级扰动靠更大攻击范围移除水印。这篇工作评估推理时 LLM 水印在黑盒、有限查询下的移除难度，并比较字符级、token 级和句子级编辑。
You have reached the end