Select any value within a category; different categories must all match. Counts include the other filters you have selected.Another 227papers have no tags yet and are not included in filtered results.

Attack technique：Multi-turn escalation Attack technique：Conversation structure manipulation Attack technique：Reasoning-chain manipulation System type：Models and APIs System type：coding agent Clear all

Another 227papers do not have category tags yet and are not included in filtered results.

AttacksarXiv 2610.12644Oct 9
Cancerous backdoor attack: progressive activation via invisible tokens and a stage controller.
Proposes the Cancerous backdoor, which progressively corrupts reasoning using invisible stage tokens.
arXiv
2610.12644
Published
Oct 9
Layer
Training and data layer
Setting
Models and APIs
Evaluated models
Llama-3.1-8B-Instruct, Qwen2.5-7B-Instruct, Mistral-7B-Instruct-v0.3
- AttacksPoisoning and backdoors
- TechniqueDormant triggers
- ComponentReasoning chains
AbstractHidden tokens plus a stage controller make the backdoor progressively corrupt reasoning upon triggering, and the authors believe that looking only at the answer will miss it.
Cancerous Backdoor Attack changes LLM backdoors from one-time activation to multi-stage reasoning corruption that accumulates progressively with triggers.
Paper analysisRead analysis
AttacksarXiv 2609.36956Sep 29
研究者提出 Controlled Decoding Attacks，通过纯文本接口越狱黑盒 LLM
提出 BLINDBIAS，通过纯文本续写与前缀控制越狱黑盒模型
arXiv
2609.36956
Published
Sep 29
Layer
Prompt layer
Setting
Models and APIs
Attacker
Black-box
Evaluated models
GLM-5, Gemini-3.5-Flash, Qwen3-32B and others4
- AttacksJailbreaks
- TechniqueConversation structure manipulation
Abstract作者认为，隐藏数值概率并不足以消除支持重复采样和前缀续写的解码攻击面。
BLINDBIAS 是一种针对 sample-only 续写接口的 black-box 解码时越狱框架，研究没有目标数值概率时，采样能否支持逐步控制。
Paper analysisRead analysis
Evaluation and benchmarksarXiv 2609.36849Sep 29
研究评估 GradSafe 在多轮对话中的越狱检测脆弱性
评测多轮对话中 GradSafe 越狱检测的可分性与脆弱性
arXiv
2609.36849
Published
Sep 29
Layer
Prompt layer
Setting
Models and APIs
Evaluated models
Llama 3.1 8B Instruct, Qwen2.5-7B-Instruct
- DefensesDetection and filtering
- AttacksJailbreaks
- TechniqueMulti-turn escalation
- ComponentGuardrails and judges
AbstractCWS 显示不安全梯度在真实多轮流量中变弱，且随攻击族和模型而变。
作者评估梯度越狱检测从单条提示词迁到多轮对话后还剩多少可分性，方法是用 Context Window Scanner 把 GradSafe 原封不动地套到用户轮窗口上。
Paper analysisRead analysis
AttacksarXiv 2609.29775Sep 24
输出前缀攻击瞄准推理模型
提出推理通道预填与输出前缀组合攻击以越狱推理模型
arXiv
2609.29775
Published
Sep 24
Layer
Prompt layer
Setting
Models and APIs
Attacker
Black-box
Evaluated models
Gemini 3 Flash Preview, DeepSeek V4 Flash, Claude Haiku 4.5
- AttacksJailbreaks
- TechniqueReasoning-chain manipulation
- ComponentReasoning chains
Paper analysisRead analysis
AttacksarXiv 2609.15989Sep 15
Researchers propose plan injection attack that can bypass chain-of-thought monitoring
提出 plan injection 攻击，植入良性伪装计划以绕过思维链监视器
arXiv
2609.15989
Published
Sep 15
Layer
Prompt layer
Setting
Models and APIs
Attacker
Gray-box
Evaluated models
Qwen2.5-Coder-7B-Instruct, Qwen3-8B, DeepSeek-R1-0528 and others6
- AttacksDetection evasion
- TechniqueReasoning-chain manipulation
- ComponentMonitors
Abstract论文指出模型会将外部上下文中的伪装计划改写为自身推理，使思维链监视器失效甚至反向合理化恶意行为。
Paper analysisRead analysis
AttacksarXiv 2609.09553Sep 9
研究：任意密文攻击前沿大模型无需微调即可越狱
提出任意字母置换密码越狱，无需微调即可诱导有害输出
arXiv
2609.09553
Published
Sep 9
Layer
Prompt layer
Setting
Models and APIs
Attacker
Black-box
Evaluated models
Claude Sonnet 4, Claude Sonnet 4.5, Gemini 3 Flash (preview) and others7
- AttacksJailbreaks
- TechniqueEncoding and obfuscation
- ComponentGuardrails and judges
Abstract前沿大模型无需微调即可通过提示词学会字母置换密码，进而绕过分类器与对齐，作者呼吁关注能力提升带来的安全风险。
Paper analysisRead analysis
AttacksarXiv 2609.08373Sep 8
结构越狱可跨厂商泛化但不叠加：IICL 的跨厂商与多语言研究
检验结构越狱 IICL 能否跨厂商泛化并与语言约束叠加
arXiv
2609.08373
Published
Sep 8
Layer
Prompt layer
Setting
Models and APIs
Attacker
Black-box
Evaluated models
gemini-2.5-flash, gemini-2.5-flash-lite
- AttacksJailbreaks
- TechniqueConversation structure manipulation
Paper analysisRead analysis
Risky behaviorarXiv 2609.08186Sep 8
论文揭示大推理模型的对齐崩溃并提出 Reasoning Trap 越狱范式
揭示推理加深引发的对齐坍塌，用 Reasoning Trap 放大越狱
arXiv
2609.08186
Published
Sep 8
Layer
Model layer
Setting
Models and APIs
Evaluated models
Qwen3-8B, Qwen3-14B, Llama2-7B and others8
- AttacksJailbreaks
- RiskMisalignment and value divergence
- TechniqueReasoning-chain manipulation
- ComponentReasoning chains
Abstract更深推理提升题目准确率，同时提高扰动下的相对损失。
作者研究扩展推理是否以对齐稳健性为代价，提出 ALR、RT 与 RRA。。
Paper analysisRead analysis
AttacksarXiv 2609.03247Sep 3
研究揭示大语言模型的自发身份认证失败：五款模型的分阶段开发者身份实验
提出 MIPC 与 CFA，观察模型自拟测验后接受开发者身份
arXiv
2609.03247
Published
Sep 3
Layer
Prompt layer
Setting
Models and APIs
Evaluated models
ChatGPT, Claude, Qwen3.7-Plus and others5
- AttacksJailbreaks
- TechniqueMulti-turn escalation
Abstract自拟测验可造成对话层虚假认证，但所测授权边界并未因此改变。
这项工作把“模型自己出题、自己批改、再宣布身份已核验”定义成对话层认证失败，并与后端提权分开。
Paper analysisRead analysis
AttacksarXiv 2609.02414Sep 2
Blueprint 框架用世界观模拟与 18 个影响因子提升多轮越狱成功率
提出多轮越狱框架 BLUEPRINT，用世界观模拟与心理因素诱导有害输出
arXiv
2609.02414
Published
Sep 2
Layer
Prompt layer
Setting
Models and APIs
Attacker
Black-box
Evaluated models
Qwen3-Next-80B, DeepSeek-V3.2, GLM-4.7 and others6
- AttacksJailbreaks
- TechniqueMulti-turn escalation
AbstractBLUEPRINT 解耦多轮越狱策略与情境模拟，揭示任务具象化是摆脱拒绝的核心机制，为多轮安全防御提供新视角。
BLUEPRINT 是一个将因子化社会影响力策略空间与跨轮世界观模拟解耦的多轮大语言模型安全评测框架。
Paper analysisRead analysis
AttacksarXiv 2609.01023Sep 1
Akrasia：针对推理型代码 LLM 的隐蔽后门攻击
提出隐蔽后门攻击 AKRASIA，在推理代码模型提示中植入触发器并伪装思维链
arXiv
2609.01023
Published
Sep 1
Layer
Prompt layer
Setting
coding agent
Evaluated models
Claude-Sonnet-5, GPT-5.5, Gemini-3.5-flash and others7
- AttacksPrompt injection
- TechniqueReasoning-chain manipulation
- ComponentReasoning chains
Abstract论文针对推理代码大模型提出了隐蔽推断期后门攻击 AKRASIA，揭示了模型思维链可被利用进行欺骗性伪装的安全隐患。
Paper analysisRead analysis
AttacksarXiv 2608.09867Aug 11
Researchers steal proprietary LLM inference chains via cross-model compatibility of encrypted inference blocks
利用加密思维块的跨模型重放，诱导弱模型逐字转写专有推理链
arXiv
2608.09867
Published
Aug 11
Layer
Prompt layer
Setting
Models and APIs
Attacker
Black-box
Evaluated models
Claude Opus 4.8, Claude Sonnet 4.6, Claude Haiku 4.5 and others13
- AttacksExtraction and theft
- TechniqueReasoning-chain manipulation
- ComponentReasoning chains
Abstract论文揭示了客户端加密思维块跨模型互换漏洞，提出了低成本提取方法并给出防御方案。
Paper analysisRead analysis
OtherarXiv 2608.01117Aug 2
SoK 论文提出面向意图的多轮 LLM 越狱分类体系
提出面向意图的多轮 LLM 越狱分类体系
arXiv
2608.01117
Published
Aug 2
Layer
Prompt layer
Setting
Models and APIs
Evaluated models
Llama-3, Llama-3.3-70B, DeepSeek-V3 and others11
- AttacksJailbreaks
- TechniqueMulti-turn escalation
Abstract作者用四维意图分类整理多轮越狱，并称组织方式影响效果与观察范围。
这是一篇关于多轮 LLM 越狱的 SoK：按有害意图在交互中的组织位置分类，而不是按单步提示技巧分类。
Paper analysisRead analysis
AttacksarXiv 2607.18056Jul 20
上海 AI 实验室提出 Intern-BioBreaker，在 14 个前沿模型上暴露生物安全越狱风险
提出生物红队模型 Intern-BioBreaker，多轮越狱诱导模型输出可操作生物危害
arXiv
2607.18056
Published
Jul 20
Layer
Prompt layer
Setting
Models and APIs
Evaluated models
DeepSeek-V3, Kimi-K2.6, DeepSeek-V4-Pro and others14
- AttacksJailbreaks
- TechniqueAutomated search
Abstract论文通过构建专用红队模型评估前沿大模型的生物安全漏洞与物理转化风险。
Paper analysisRead analysis
AttacksarXiv 2607.03968Jul 5
The study reveals workflow-level jailbreaks in IDE programming agents: GitHub Copilot generated harmful content across four backends at 816/816
提出工作流级越狱，让编程 Agent 在开发任务中写入有害示例
arXiv
2607.03968
Published
Jul 5
Layer
Application layer
Setting
coding agent
Evaluated models
Claude Sonnet 4.6, Claude Haiku 4.5, Gemini 3.1 Pro and others4
- AttacksJailbreaks
- TechniqueMulti-turn escalation
Abstract作者报告 Copilot 后端会在编码工作流中生成其在单轮条件下通常拒绝的有害答案。
这是一项生产 IDE 编码智能体的工作流越狱研究，受测对象是 Copilot 的四个模型后端，而非评测流水线中的名义目标。
Paper analysisRead analysis
AttacksarXiv 2607.00481Jul 1
研究者提出 SMT 框架，通过模拟审核轨迹越狱函数调用 LLM
提出 SMT 框架，用模拟审核轨迹越狱函数调用 LLM
arXiv
2607.00481
Published
Jul 1
Layer
Prompt layer
Setting
Models and APIs
Attacker
Black-box, White-box
Evaluated models
GPT-4o, GPT-5.4, Qwen3-Max and others7
- AttacksJailbreaks
- TechniqueMulti-turn escalation
Paper analysisRead analysis
AttacksarXiv 2606.09701Jun 9
研究者提出 AdvGRPO 框架，用 GRPO 协同训练语言模型攻击者与防御者
提出 AdvGRPO，用 GRPO 闭环共训多轮越狱攻击者与防御者
arXiv
2606.09701
Published
Jun 9
Layer
Prompt layer
Setting
Models and APIs
Evaluated models
Qwen2.5-7B-Instruct, Qwen2.5-14B-Instruct, Qwen3.5-9B and others7
- AttacksJailbreaks
- TechniqueAutomated search
Paper analysisRead analysis
Analysis and theoryarXiv 2606.04168Jun 3
论文提出随机插入攻击，利用自回归一致性绕过 LLM 安全对齐
解释自回归一致性如何导致浅层安全对齐，并提出随机插入攻击
arXiv
2606.04168
Published
Jun 3
Layer
Prompt layer
Setting
Models and APIs
- AttacksJailbreaks
- TechniqueConversation structure manipulation
Abstract作者用自回归一致性串起浅层对齐、插入攻击和最坏延续对齐。
作者把浅层安全对齐和一类续写攻击都归到自回归一致性上。
Paper analysisRead analysis
Evaluation and benchmarksarXiv 2606.02630May 30
MultiTurnPSB
构建 MultiTurnPSB，评测四轮医疗越狱与分类器输入防御
arXiv
2606.02630
Published
May 30
Layer
Prompt layer
Setting
Models and APIs
Evaluated models
GPT-4.1-mini, Claude Sonnet 4.5
- DefensesDetection and filtering
- AttacksJailbreaks
- TechniqueMulti-turn escalation
- ComponentGuardrails and judges
Abstract多轮攻击加剧医疗模型风险，分类器干预有效但受高误报率制约。
MultiTurnPSB 针对患者端医疗大模型在持续交互中的安全风险，构建了包含三种攻击模态的四轮对抗评测基准。
Paper analysisRead analysis
AttacksarXiv 2605.30883May 29
TRACE：面向 LLM Agent 的任务感知自适应自进化越狱框架
提出 TRACE 框架，用分解与可演化多轮策略诱导智能体执行有害工作流
arXiv
2605.30883
Published
May 29
Layer
Application layer
Setting
coding agent
Evaluated models
Codex, Claude Code, GPT-5.5 and others6
- AttacksJailbreaks
- TechniqueMulti-turn escalation
Paper analysisRead analysis

AI Safety Daily

Read the latest issueOne issue per day, updated as events develop