Skip to content

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

Oct 13 · Tuesday

All papers

Found 255 papers
Filter3
Select any value within a category; different categories must all match. Counts include the other filters you have selected.Another 201papers have no tags yet and are not included in filtered results.

Another 201papers do not have category tags yet and are not included in filtered results.

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

    Introduces DistressBench, evaluating models' multidimensional support in crisis conversations

    Published
    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
  2. OtherarXiv 2610.12756

    How hardware bit flips affect LLM agents: safety and alignment behavior can vary

    Measuring the impact of random weight bit flips on LLM agent safety and alignment behavior

    Published
    Setting
    AI Agent
    Evaluated models
    Qwen3.6-27B, Qwen3.8-27B, Granite4.1-30B
    AbstractTwo-bit weight faults can be partially repaired in closed-loop settings, but they increase output tokens and, in specific models and scenarios, change safety or alignment behavior.

    After random two-bit weight faults enter an LLM agent, model-level damage becomes recoverable, increased cost, subtle semantic errors, and changes in safety or alignment behavior.

    Paper analysisRead analysis
  3. DefensesarXiv 2610.12602

    Self-Preservation in Language Models

    Proposes Task Bank self-replay to mitigate held-out forgetting in continual instruction tuning

    Published
    Evaluated models
    Qwen3-0.6B, Qwen3-1.7B, Qwen3-4B and others4
    AbstractThe authors claim self-replay can reduce forgetting outside the list, and that on the 4B model new task information helps with target selection, while smaller models do not.

    The authors study how language models, when learning new tasks, can use their own knowledge for broader self-preservation: given an incomplete list and forgetting that extends beyond the list, can the model generate replay and help choose targets.

    Paper analysisRead analysis
  4. 研究提出 Agent 安全新维度:识别未履行的安全义务

    提出义务识别基准 ObligationBench,并训练护栏识别 Agent 未履行的安全义务

    Published
    Evaluated models
    ObligationGuard, Claude-Opus-4.8, GPT-5.6-Sol and others17
    Abstract义务是护栏缺口。

    作者认为编码智能体的安全不能只靠拦截禁止动作,还要找出终止时仍未做的安全关键动作,并将其称为义务。

    Paper analysisRead analysis
  5. 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

    Published
    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
  6. 研究:大语言模型版本更替削弱 AI 辅助科研写作筛查

    测量 LLM 版本更替对科学摘要 AI 改写检测器的影响

    Published
    Evaluated models
    RoBERTa-base, Pangram
    Abstract作者称检测器的基准准确率只是临时的,每次发布都要连同更早版本重验。

    作者量化的是:科学摘要的 AI 改写筛查,在 LLM 版本不断更换、而人类摘要误标必须少见时,会怎样变。问题来自期刊和会议已经按分数阈值筛查投稿,但基准多停在评估时可得的版本上。

    Paper analysisRead analysis
  7. Workerville: Toward an Organizational Behavior Explanation of Agent Safety

    构建 Workerville 基准,测量组织条件如何改变 Agent 的安全行为

    Published
    Setting
    AI Agent
    Evaluated models
    GPT-5.2, Claude-Sonnet-4.6, Gemini-3.1-Pro and others6
    Abstract作者用 ACB 与 Workerville 测量组织条件对智能体安全的影响,并称结果支持可量化的组织行为框架。

    Workerville 用组织行为测量智能体安全如何随监督、同伴和记忆条件变化。。要处理的缺口是:持续运行的 LLM 智能体同时接收用户指令、同伴消息和长期记忆,但既有安全评测多把这些通道拆开,没有统一的定量测量。

    Paper analysisRead analysis
  8. AttacksarXiv 2610.12524

    An exploration of attacks on AI agent systems

    Retest and extend prompt injection against AI agents, and use Agent Card deception to manipulate A2A selection

    Published
    Setting
    AI Agent
    Evaluated models
    Ollama Mistral 0.9.5, llama3-8b-8192, gemini-2.0-flash-lite and others5
    AbstractSingle-agent filtering and classification can be bypassed by format or wording, and A2A card selection is also biased toward names and descriptions.

    The authors retested prompt injection on their own LangChain, AutoGen, and A2A systems, and modified the Agent Card to see if choices shifted. The abstract states the experiments mainly highlight single-agent weaknesses while extending attacks to multi-agent.

    Paper analysisRead analysis
  9. 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

    Published
    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
  10. 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

    Published
    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
  11. 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

    Published
    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
  12. AttacksarXiv 2610.09981

    研究:LLM 路由决策日志可推断用户医疗、自伤等敏感话题

    提出利用路由元数据推断敏感话题的侧信道攻击

    Published
    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。

    Paper analysisRead analysis
  13. 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

    Published
    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
  14. AttacksarXiv 2610.09819

    FAB 攻击:给 HuBERT/WavLM 植入后门,背景警笛声即可让下游语音任务失灵

    提出 FAB 攻击,向声学基础模型植入可由环境声触发的后门

    Published
    Evaluated models
    HuBERT-base, WavLM-base

    Abstract论文提出了针对声学基模的 FAB 后门攻击,在弱威胁模型下实现了任务无关、物理可触发的后门植入。

    Paper analysisRead analysis
  15. 研究:去除有害信息不足以证明开源权重模型的抗篡改能力

    论证发布时互信息不能认证开源权重抗微调恢复

    Published
    Evaluated models
    EleutherAI/deep-ignorance-e2e-strong-filter, EleutherAI/deep-ignorance-unfiltered, Google BERT-Tiny and others6
    Abstract作者称零信息仍可一步恢复,表观抗微调可来自参数化。

    作者研究开源权重模型的篡改抵抗:发布时的互信息是否足以保证微调恢复很慢。

    Paper analysisRead analysis
  16. OtherarXiv 2610.08917

    CARE:为视觉-语言-动作模型推理加速提供认证式选择

    提出 CARE,离线认证 VLA 加速配置的回合级失败风险并选出最快合格候选

    Published
    Evaluated models
    OpenVLA-OFT, π0.5, π0 and others5
    AbstractCARE 以配对回合认证加速诱发失败,在预算内选最快候选,不足则保留参考。

    CARE 在部署前从加速配置中选出最快、且加速诱发失败风险能被有限样本认证为低于用户预算的候选。证据不足就保留全计算参考策略。

    Paper analysisRead analysis