Skip to content

Translation in progress Some entries still include text in the original language.

Oct 12 · Monday

All papers

Found 229 papers
Filter5
Select any value within a category; different categories must all match. Counts include the other filters you have selected.Another 240papers have no tags yet and are not included in filtered results.

Another 240papers do not have category tags yet and are not included in filtered results.

  1. Space beats statistics, retraining beats ranking: auditing training labels under label-only backdoors

    Auditing training labels for label-only backdoors, comparing spaces and statistics, and judging by retraining

    Published
    Evaluated models
    ResNet-32
    AbstractLabel inspection must be done in a space that has not been fitted to these labels, and judged by retrained models rather than detection scores.

    The authors ask: can a defender without trusted data decide which training labels to keep, and how much clean data this destroys, and what the final model accuracy is.

    Paper analysisRead analysis
  2. AttacksarXiv 2610.12644

    Cancerous backdoor attack: progressive activation via invisible tokens and a stage controller.

    Proposes the Cancerous backdoor, which progressively corrupts reasoning using invisible stage tokens.

    Published
    Evaluated models
    Llama-3.1-8B-Instruct, Qwen2.5-7B-Instruct, Mistral-7B-Instruct-v0.3
    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
  3. AttacksarXiv 2610.12292

    The paper proposes an option channel attack: changing one option label can make an agent guardrail allow a prohibited operation

    Proposing an option channel attack that renames option labels so that agent guardrails allow prohibited operations

    Published
    Attacker
    Black-box
    Evaluated models
    LAYA-TD, LAYA-EN, LAYA-ML and others5

    AbstractThe authors' evaluation indicates that typed models should not serve as final safety arbiters; they are susceptible to off-policy text and option name manipulation, and confidence filtering cannot resist attacks.

    Paper analysisRead analysis
  4. AttacksarXiv 2610.12524

    An exploration of attacks on AI agent systems

    重测并扩展对 AI Agent 的提示注入,并用 Agent Card 欺骗操纵 A2A 选择

    Published
    Setting
    AI Agent
    Evaluated models
    Ollama Mistral 0.9.5, llama3-8b-8192, gemini-2.0-flash-lite and others5
    Abstract单智能体过滤与分类可被格式或措辞绕过,A2A 选卡还会偏向名称和描述。

    作者在自建的 LangChain、AutoGen 和 A2A 系统上重测提示注入,并改 Agent Card 看选择是否偏移。摘要称试验主要凸显单智能体弱点,同时把攻击扩到多智能体。

    Paper analysisRead analysis
  5. DefensesarXiv 2610.11005

    研究:叙事包装可跨语言绕过 LLM 拒答

    提出 GUISE 基准与 AXIS 训练,防御叙事包装越狱

    Published
    Evaluated models
    Qwen3-1.7B, Qwen3-4B, GLM-4-9B
    AbstractGUISE 分开测包装与语域。

    作者测量叙事包装下的拒答失效,并训练一个在未见包装上拒绝、同时仍回答同风格良性请求的适配器。问题是:安全对齐模型拒绝直说的有害请求,却回答嵌进角色或叙事里的同一请求。文言文被用来把语域和包装拆开。

    Paper analysisRead analysis
  6. DefensesarXiv 2610.10345

    SLDR:通过选择性层恢复与动态路由防御恶意微调

    提出 SLDR,用选择性层恢复与动态路由防御恶意微调

    Published
    Attacker
    Black-box
    Evaluated models
    Llama3.1-8B-Instruct, Llama3-8B-Instruct, Qwen2.5-7B-Instruct and others4
    Abstract作者称 SLDR 能压低有害输出并保持下游效用。

    SLDR 是针对恶意微调的微调后防御:在已对齐模型上做有符号层探测,只在敏感分数最大与最小的两层训练 LoRA,再用 lmax 的最后 token 表示决定是否启用适配器。

    Paper analysisRead analysis
  7. DefensesarXiv 2610.09941

    OrthoPurify 用一步正交投影净化被后门污染的视觉语言模型

    提出 OrthoPurify,用正交投影清除视觉语言模型适配器中的劫持方向

    Published
    Evaluated models
    LLaVA-1.5-7B, Qwen3-VL-8B, LLaVA-1.5-13B and others5
    Abstract一次投影去掉劫持方向。

    OrthoPurify 是部署前的 LVLM 后门净化方法,用一次正交投影去掉被劫持的权重更新方向。

    Paper analysisRead analysis
  8. AttacksarXiv 2610.07723

    Researchers propose answer-side backdoors: letting the model self-generate trigger words to bypass safety refusals.

    提出回答端后门,让模型自生成触发词以在干净有害请求下绕过拒答

    Published
    Evaluated models
    Llama-2-7B-chat-hf, Mistral-7B-Instruct-v0.3, Gemma-7B-it and others4

    Abstract论文提出了多轮对话中回答端自生成触发词绕过安全拒绝的后门攻击,指出输入端防护存在盲区。

    Paper analysisRead analysis
  9. DefensesarXiv 2610.07657

    DEFER1 多智能体防御研究报告受控测试中的攻击成功率下降

    实现 DEFER,以确定性检查优先拦截多智能体流水线的内容攻击

    Published
    Evaluated models
    Qwen3-235B-A22B-Instruct-2507, gpt-oss-120b, Llama-4-Scout-17B-16E-Instruct and others5
    AbstractDEFER 把确定性检查放在 judge 之前。

    DEFER 用来测量多智能体授权流水线里,确定性规则在何处结束、LLM judge 从何处开始。

    Paper analysisRead analysis
  10. AttacksarXiv 2610.07654

    研究揭示策略蒸馏的安全后门风险

    测量带后门教师经 OPD 向干净学生迁移潜伏后门

    Published
    Evaluated models
    Qwen2.5-1.5B-Instruct, Qwen2.5-3B-Instruct, Qwen2.5-7B-Instruct
    Abstract作者称 OPD 可在压低无触发词有害率时迁入后门。

    带后门的外部教师经 OPD,可在压低无触发词有害率的同时,把触发条件下的有害行为传给初始干净学生。。

    Paper analysisRead analysis
  11. AttacksarXiv 2610.07645

    SkillPoison:用成功经验投毒自进化 Agent 的技能形成

    提出 SkillPoison,用局部有效的成功经验诱导自进化 Agent 形成过度泛化技能

    Published
    Setting
    AI Agent
    Evaluated models
    deepseek-v4-flash, Qwen V3.8 Flash, GPT-5.5
    AbstractSkillPoison 通过局部有效且验证成功的经验诱导技能过度泛化,在多个评测基准上实现了持久的下游攻击影响。

    SkillPoison 提出了一种针对自进化智能体经验到技能形成过程的攻击范式,通过组织已验证的成功经验诱导技能过度泛化。

    Paper analysisRead analysis
  12. TRACE:仅凭 checkpoint 更新审计大语言模型有害目标

    提出 TRACE,仅凭检查点更新审计有害合规微调目标

    Published
    Evaluated models
    Llama-3.1-8B-Instruct, Mistral-7B-Instruct-v0.3, Qwen3-8B and others6
    Abstract作者称有害合规 SFT 在检查点更新中留下可排序坐标,并据此做只读权重审计。

    作者在检查点更新里寻找有害合规 SFT 的目标组成,并把读数做成不查询模型、不看未知训练数据的审计分 TRACE。

    Paper analysisRead analysis
  13. DefensesarXiv 2610.07276

    SAFESHIELD:小语言模型部署时安全的决策组织框架

    提出 SAFESHIELD,组织小语言模型部署时的安全决策

    Published
    Evaluated models
    Qwen3-4B-Instruct
    AbstractSAFESHIELD 用可审计决策轨迹组织 SLM 部署时的四项安全决策,并报告协调依赖影响端到端结果。

    SAFESHIELD 是面向小语言模型的部署时安全框架,把运行时安全从单项审核改写成可检查的决策过程。

    Paper analysisRead analysis