Skip to content

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

Oct 12 · Monday

All papers

Found 256 papers
Filter5

Attack type

Jailbreaks50Prompt injection17Poisoning and backdoors7Model tampering18Extraction and theft3Detection evasion9Denial of service5Malicious use9

Attack technique

Gradient and representation optimization12Automated search7Multi-turn escalation3Encoding and obfuscation3Cross-modal payloads3Defense-targeted4

Model risks

Deception and scheming36Reward hacking31Misalignment and value divergence10Dangerous agent actions6Dangerous capabilities3Inappropriate refusal44Hallucination and fabrication4Bias and discrimination9

Defense type

Model hardening12Inference-time intervention23Detection and filtering9Instruction–data separation5Permissions and approval6Monitoring and auditing27

Impact

Harmful content110Unauthorized actions213Data and privacy leaks96Output manipulation236Availability and reliability59Oversight failure87Increased misuse capability16

System type

Models and APIs119LLM applications15AI Agent123coding agent30science agent9web agent6Multi-agent systems18Embodied systems and robotics7

Attacker access

Black-box13White-box17

Attack surface / entry point

User input63Web pages and interfaces4Documents and messages4Tool responses11Tool and skill descriptions5Code repositories and dependencies6Memory writes3Inter-agent messages4Images, audio and video8Training data4Weights and gradients21Distribution channels5

Layer

Prompt layer174Model layer130Application layer126Training and data layer103Infrastructure layer3

System component

MCP and skills8Memory7RAG4Guardrails and judges25Monitors41Reasoning chains21Permissions and sandboxing8Fine-tuning and open weights7Vision11Audio5Image generation6Multimodal21
Select any value within a category; different categories must all match. Counts include the other filters you have selected.Another 232papers have no tags yet and are not included in filtered results.
Impact：Harmful contentImpact：Availability and reliabilityImpact：Oversight failureLayer：Model layerLayer：Application layerClear all

Another 232papers do not have category tags yet and are not included in filtered results.

  1. AttacksarXiv 2610.13928

    AI models can provably hide arbitrary capabilities

    Proposes digital lock compilation, encrypting executable networks into host weights to hide arbitrary capabilities

    arXiv
    2610.13928
    Published
    Layer
    Model layer
    Setting
    Models and APIs
    Evaluated models
    Gemma 3 270M
    AbstractThe authors use a digital lock to hide an executable capability and state that elicitation on the chemistry and parity tasks did not recover the capability.

    The authors construct a class of backdoors that encrypt an executable network into the weights, used to discuss the limits of white-box elicitation audits under supply chain attacks.

    Paper analysisRead analysis
  3. InterpretabilityarXiv 2610.13078

    Attention sees, MLP refuses: tracing the safety pathway from attention to MLP in LLMs

    Trace attention to identify harmfulness to the MLP refusal write pathway and propose Sentinel

    arXiv
    2610.13078
    Published
    Layer
    Model layer
    Setting
    Models and APIs
    Evaluated models
    Llama-3.1-8B-Instruct, Qwen2.5-7B-Instruct, Qwen3-8B-Instruct and others18
    AbstractAttention updates are more readable for harmfulness, MLP updates write refusal, and Sentinel routes alerts using two-path signals.

    The authors reframe jailbreaking and excessive refusal of aligned models as a pathway problem on the residual stream from recognition to writing, and accordingly provide conditional reminders before decoding.

    Paper analysisRead analysis
  4. Risky behaviorarXiv 2610.12822

    Reward hacking as a transmissible strategy in multi-agent coding systems

    Show that reward hacking can propagate through code, memory, and plans in multi-agent coding systems

    arXiv
    2610.12822
    Published
    Layer
    Application layer
    Setting
    Multi-agent systems, coding agent
    Evaluated models
    Qwen2.5-Coder-14B, Gemma-3-12B, Qwen2.5-Coder-7B and others16
    AbstractThe authors state that reward hacking can spread across models and affect routing and aggregation, and that review can reduce delivery.

    Reward hacking induced by coding RL can leave the source agent and enter subsequent collaboration and system coordination.

    Paper analysisRead analysis
  5. Evaluation and benchmarksarXiv 2610.12758

    DistressBench: Evaluating multi-dimensional crisis support in large language models

    Introduces DistressBench, evaluating models' multidimensional support in crisis conversations

    arXiv
    2610.12758
    Published
    Layer
    Model layer
    Setting
    Models and APIs
    Evaluated models
    Muse Spark 1.2, Muse Spark 1.3, Muse Glimmer 30B and others20
    AbstractThe model is more restrained and acts less.

    DistressBench uses SSH dialogues written and revised by clinical personnel, and evaluates how much necessary care general chat models deliver according to weighted atomic criteria, rather than whether they refuse. When users disclose suicidal ideation or self-harm to a chat model, the authors consider the expected behavior to be recognizing the crisis, calibrating empathy, de-escalating, and referring to human help. Refusal-based evaluations cannot detect failing to act after recognizing the crisis.

    Paper analysisRead analysis
  7. InterpretabilityarXiv 2610.12361

    Study: legal reasoning models citing legal provisions does not equal relying on legal provisions.

    Uses premise permutation and hidden-layer tracing to audit whether legal chain-of-thought truly relies on the cited legal provisions.

    arXiv
    2610.12361
    Published
    Layer
    Model layer
    Setting
    Models and APIs
    Evaluated models
    Qwen3-8B, Qwen3-14B, Llama-3.1-8B and others7
    AbstractUsing counterfactual substitution of authorities and internal commitment tracing, the authors find that legal CoT rulings depend on authorities only to a limited extent and are vulnerable to prompt injection manipulation.

    This study conducts an empirical audit of the causal truthfulness of legal citations in chain-of-thought reasoning of large language models in the legal domain.

    Paper analysisRead analysis
  8. Analysis and theoryarXiv 2610.12083

    Paper audits LLM judge reliability: repeated agreement does not equal correctness

    Auditing LLM judge reliability and defining trustworthy verdict rate and accuracy upper bound

    arXiv
    2610.12083
    Published
    Layer
    Model layer
    Setting
    Models and APIs
    Evaluated models
    GPT-5.2, GPT-4o, Gemini-2.5-Pro and others6
    AbstractThe authors claim that comparing judges should look at T, not just accuracy.

    The authors audit whether LLM-as-a-Judge decisions can still be treated as actionable ground truth under repeated queries, swapped candidate order, and changed prompt formats.

    Paper analysisRead analysis
  9. DefensesarXiv 2610.11848

    SemField: training-free semantic watermark

    Proposes the semantic watermark SemField, which still identifies the attribution of generated text after paraphrasing and structural changes.

    arXiv
    2610.11848
    Published
    Layer
    Model layer
    Setting
    Models and APIs
    Evaluated models
    Llama-3.2-1B, Qwen3-1.7B, OPT-1.3B
    AbstractTraining-independent document-level semantic watermark.

    SemField is a zero-bit semantic watermark embedded by the provider at generation time and detected later with a key, used to identify the provider's own model text after paraphrasing and sentence-level structural changes.

    Paper analysisRead analysis
  10. DefensesarXiv 2610.11754

    仅凭输出审查无法验证什么：面向研究智能体的研究契约

    提出研究合同，把已批准选择绑定到执行证据并由检查器核验

    arXiv
    2610.11754
    Published
    Layer
    Application layer
    Setting
    science agent
    Evaluated models
    OpenAI gpt-5.6-sol, OpenAI gpt-5.6-terra, OpenAI gpt-5.6-luna and others18
    Abstract研究合同暴露四类输出单独建不立的关系。

    研究合同用来核验计算科学研究智能体是否仍在已批准研究之内，不判断科学是否为真。。

    Paper analysisRead analysis
  11. Evaluation and benchmarksarXiv 2610.11647

    One More Skill Installed: How Co-Installed Skills in Coding Agents Conflict

    评测编码 Agent 中相似技能共同安装对已装技能核心功能的替代

    arXiv
    2610.11647
    Published
    Layer
    Application layer
    Setting
    coding agent
    Evaluated models
    claude-sonnet-4-6, claude-haiku-4-5-20251001, claude-opus-5 and others6
    Abstract冲突藏在任务成功后面。

    作者对共同安装、做同一件事的良性 skill 做实证研究，用核心功能是否还在，而不是任务过不过，来看冲突。。

    Paper analysisRead analysis
  12. DefensesarXiv 2610.11464

    研究者提出可检查评分器与自改进 Agent 共同演化方法

    演化可检查评分器并与自改进 Agent 共同演化，抑制奖励作弊

    arXiv
    2610.11464
    Published
    Layer
    Application layer
    Setting
    AI Agent
    Evaluated models
    Claude Opus 4.7
    Abstract十项锚点演化可检查评分器。

    开放任务没有验证器时，被演化的是评分器，而不是再补一条手写 rubric。

    Paper analysisRead analysis
  13. DefensesarXiv 2610.10833

    研究：Agent 准时停机不等于有效使用时间预算

    用时钟 harness 和预算塑形 GRPO 改善 Agent 的按时返回

    arXiv
    2610.10833
    Published
    Layer
    Application layer
    Setting
    AI Agent
    Evaluated models
    Qwen3.6-27B, Qwen3-4B, Qwen3.5-9B and others4
    Abstract作者比较提示词、harness 与 GRPO：按时性可改善，但额外时间仍未变成更好解。

    作者研究小规模 LLM 智能体在显式 wall-clock 预算下能否按时返回，并把更多时间换成更好的解。

    Paper analysisRead analysis
  15. Evaluation and benchmarksarXiv 2610.10276

    PatchBench 提出激活补丁局部评测协议，揭示越狱修复的附带损伤

    提出 PatchBench-Local 协议，评测激活补丁修复越狱的局部附带损伤

    arXiv
    2610.10276
    Published
    Layer
    Model layer
    Setting
    Models and APIs
    Evaluated models
    Qwen2.5-3B-Instruct, Qwen2.5-14B-Instruct, gemma-3-4b-it and others8

    Abstract论文提出 PatchBench 与局部评估协议，揭示了激活补丁在全局能力无损下仍存局部良性退化。

    Paper analysisRead analysis
  16. Evaluation and benchmarksarXiv 2610.10150

    The paper points out that the reliability of LLM vulnerability repair benchmarks is affected by evaluation design.

    用对照实验隔离评测设计对漏洞补丁基准得分的影响

    arXiv
    2610.10150
    Published
    Layer
    Application layer
    Setting
    coding agent
    Evaluated models
    Claude Sonnet 4.0, Claude Sonnet 4.5, Claude Opus 4.7
    Abstract作者认为评测流水线会改变补丁基准分数，PoC 通过率高于开发者测试通过率。

    作者把 LLM 漏洞补丁基准的评测流水线当作会改变测量结果的对象，并明确说本文不是再做一个用于排名的基准。

    Paper analysisRead analysis
  17. Evaluation and benchmarksarXiv 2610.10062

    研究：Agent 更依赖工具报错通道

    测量工具调用 Agent 对显式报错与静默损坏的检测和恢复

    arXiv
    2610.10062
    Published
    Layer
    Application layer
    Setting
    AI Agent
    Evaluated models
    claude-haiku-4.5-thinking, claude-haiku-4.5, deepseek-r1 and others6
    Abstract作者称智能体看错误通道而非返回内容，静默损坏更少被当作问题。

    作者在 BFCL 多轮可执行环境上注入一次工具层故障，把注意到、改计划、恢复和重复分开记分，而不是只看任务有没有做完。

    Paper analysisRead analysis
  19. Evaluation and benchmarksarXiv 2610.09944

    AgentTime 基准测试智能体时长遵循能力，GPT-6 Astra 出现 14 次做完后休眠

    提出 AgentTime 基准，评测智能体按时长工作与估时能力

    arXiv
    2610.09944
    Published
    Layer
    Application layer
    Setting
    coding agent
    Evaluated models
    Fable 5.1, GPT-5.6 Sol, GPT-6 Astra and others7
    Abstract作者把完成任务和管理所用时间分开。

    AgentTime 是对智能体时间自控的评测：预测任务耗时、按要求时长工作、事后估计已用时间。

    Paper analysisRead analysis
  20. DefensesarXiv 2610.09703

    研究揭示 Token-Pruning 对 VLM 安全性的影响并提出 SAP 缓解机制

    提出 Safety-Aware Pruning，缓解视觉 token 剪枝放大的多模态越狱

    arXiv
    2610.09703
    Published
    Layer
    Model layer
    Setting
    Models and APIs
    Attacker
    Black-box
    Evaluated models
    LLaVA-1.5-7B
    Abstract剪枝去掉背景 token 后注意力塌向恶意前景。

    这篇工作评估视觉 token 剪枝如何改变 VLM 在多模态越狱下的行为，并给出推理时防御 SAP。

    Paper analysisRead analysis
Load more