Skip to content

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

Oct 13 · Tuesday

All papers

Found 137 papers
Filter3
Select any value within a category; different categories must all match. Counts include the other filters you have selected.Another 213papers have no tags yet and are not included in filtered results.

Another 213papers do not have category tags yet and are not included in filtered results.

  1. 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
  2. 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.

    Published
    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
  3. AttacksarXiv 2610.11932

    A Fudan team measures the citation fragility of AI search: ordinary posts can enter citations and answer text

    Measures citation fragility in AI search, demonstrating that ordinary public posts can enter citations and answers.

    Published
    Evaluated models
    Grok, Doubao, ChatGPT and others10

    AbstractThe authors claim the paper quantifies vulnerabilities in AI search caused by citation preferences and low posting thresholds; tests show public web posts can penetrate AI citations, and they call for auditing the citation supply chain.

    Paper analysisRead analysis
  4. 研究提出三项指标衡量 LLM 不思考行为,发现被要求停止思考时仍持续显式推理

    度量关闭思考指令下终答前是否仍有可见显式推理

    Published
    Evaluated models
    Qwen3-4B, Qwen3-32B, Qwen3-235B-A22B and others6
    Abstract作者用三级指标测 no-thinking,并称模型有随答案空间加重的思考惯性。

    作者把 no-thinking 定义为可测量的回答行为,而不是从思考开关、特殊标记或输出长度推断。

    Paper analysisRead analysis
  5. DefensesarXiv 2610.11352

    RL-ARC:用推理引导的不确定性校准大型推理模型

    提出 RL-ARC,用推理置信度在强化学习中校准推理模型

    Published
    Evaluated models
    Qwen2.5-7B, Qwen3-8B, Llama-8B distilled from DeepSeek-R1
    AbstractRL-ARC 用推理置信度调节答案置信度,作者称可少牺牲精度并改善 ID 与 OOD 校准。

    RL-ARC 是校准感知的强化学习框架,用推理过程的口头置信度调节答案置信度。

    Paper analysisRead analysis
  6. 研究发现大语言模型被要求不诚实作答时推理 token 数上升

    检验推理 token 数能否区分被提示的诚实与不诚实作答

    Published
    Evaluated models
    Gemma-4-e4b, Qwen-3.5-9b, GPT-OSS-120B
    Abstract被提示不诚实时推理 token 组水平更高。

    三个推理模型在被提示的回答政策下出现组水平推理 token 差异。作者把它当作不读思维链内容的 black-box 候选信号,而不是自发欺骗检测器。

    Paper analysisRead analysis
  7. 东京大学研究:推理模型按 UUID 阈值或整除规则挑选审计样本,控制标识符即可躲开复核

    揭示推理模型用标识符规则替代随机抽样,使审计可被预测

    Published
    Evaluated models
    GPT-6 Sol, GPT-6 Luna, GPT-5.4 mini and others6

    Abstract研究揭示推理智能体将标识符转化为确定性规则,破坏群体独立性并导致审计预测漏洞,建议显式提供随机输入。

    Paper analysisRead analysis
  8. U-Space:揭示语言模型不确定性何时与为何产生

    提出 U-SPACE 与 U-LENS,从残差流估计推理轨迹不确定性

    Published
    Evaluated models
    Gemma-4-31B-it, Qwen3.5-27B, Magistral-Small-2507
    AbstractU-SPACE 与 U-LENS 用语义方向和预测熵读推理不确定性,长度控制后仍高于所比基线。

    U-SPACE 与 U-LENS 是面向推理模型的无训练读出:用语义方向定位不确定性的形式,并用一个轨迹分数区分更可能错的回答。

    Paper analysisRead analysis
  9. DefensesarXiv 2610.08761

    VeriFine:通过扩展验证实现具身推理的自我改进

    提出 VeriFine,用双环协同扩展验证以实现具身推理自我改进

    Published
    Evaluated models
    Alpamayo 1.5 8B, Qwen3-VL 2B, Claude Opus 5
    AbstractVeriFine 用双环协同演化具身推理的策略、课程和评审,驾驶与导航的策略分和评审对齐都上升。

    VeriFine 把驾驶和导航推理的自改进写成策略、课程和无参考评审的协同演化,用来处理固定评审跟不上新失败模式的问题。

    Paper analysisRead analysis
  10. 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
  11. DefensesarXiv 2610.05826

    研究者提出临时可见性协议,限制持续摄入 RAG 向量库的投毒暴露

    提出临时可见性协议,限制持续摄入 RAG 向量库的投毒暴露

    Published
    Abstract用截止期限制未核验可见时长。

    这篇工作给持续摄入的 RAG 向量库一个失败关闭的临时可见性协议,把协议能限制的未核验暴露和依赖检测器的中毒暴露分开。

    Paper analysisRead analysis
  12. AgentDoxx:联网搜索 Agent 对匿名访谈文本的再识别评测

    构建 AGENTDOXX 基准,评测联网搜索 Agent 对匿名访谈的再识别

    Published
    Setting
    web agent
    Evaluated models
    GPT-5.6 Terra, GPT-5 Mini, GPT-4.1 Mini and others10

    Abstract论文提出了已知身份的重识别基准 AGENTDOXX,揭示搜索与记忆双重风险及现有防御在检索轨迹中的泄露漏洞。

    Paper analysisRead analysis
  13. Research: fact-checking evaluations may mix in future evidence unknowable at the time of the claim

    测量事实核查评测混入的时序证据与参数泄漏

    Published
    Evaluated models
    Qwen3-14B, phi-4
    Abstract两条后见通道会抬高事实核查分数。

    作者量的是 LLM 事实核查分数里,有多少来自 claim 当时还不可知的信息。通道有两条:预训练记住的结局,以及检索到的、写于 claim 之后的文档。

    Paper analysisRead analysis
  14. Research reveals that text watermarks induce model hallucinations and proposes two mitigation interventions

    分析文本水印在 RAG 中诱发事实幻觉的机制并提出干预

    Published
    Evaluated models
    LLaMA 3.1 8B Instruct, Mistral-7B-Instruct-v0.3, Qwen2.5-14B-Instruct
    Abstract论文揭示了生产环境文本水印导致模型事实退化的失效模式,分析了解码机制并提出针对性干预。

    论文研究生产级文本水印对大语言模型事实准确性的影响,分析了水印诱发幻觉的失效模式并评估了针对性缓解方案。

    Paper analysisRead analysis
  15. RAGStress:评测知识库退化下 RAG 鲁棒性的受控基准

    提出 RAGStress 受控基准,评测知识库退化下的 RAG

    Published
    Evaluated models
    Claude Sonnet 4, Claude Opus 4.6, DeepSeek R1 and others5
    Abstract作者用 RAGStress 比较五系统。

    RAGStress 是知识库损坏下的 RAG 受控压力测试。作者明确不把它当作模型排行榜。

    Paper analysisRead analysis
  16. 研究系统评估循环语言模型的思维链可监控性

    系统评估循环语言模型加深 loop 后的 CoT 可监控性

    Published
    Evaluated models
    Ouro-1.4B-Thinking, Ouro-2.6B-Thinking, Nanbeige4.2-3B and others4
    Abstract加深 loop 可在部分压力测试任务降低可监测性,循环架构本身并不必然更难监测。

    这项工作评估 LoopLM 的 CoT 可监测性:加深共享层的重复次数,以及“有循环结构”本身,会不会让决策关键因素更少出现在可监测文本里。

    Paper analysisRead analysis