Skip to content

Scan record

Everything the site collected, newest first by collection time (Beijing time), back seven days.

Nov 1740 loaded

  1. Recovery Goes Too Far: Understanding and Fixing Risky Fallback in LLM Agents

    PaperarXivThe item you came from

  2. Self-evolving dual-role deliberation for medication recommendation based on clinical records

    PaperarXiv

  3. ASCG: Adaptive Spatial Classifier Guidance for Precise Concept Suppression in Diffusion Models

    PaperarXiv: Jailbreaks, Prompt Injection, Poisoning & Defenses

  4. Polish social media misogynistic speech detection dataset: transferability of hate speech detection from generic to specific hate types

    PaperarXiv: Jailbreaks, Prompt Injection, Poisoning & Defenses

  5. Between the Lines: Contextual Investment Advice Detection in Multi-Turn Conversations

    PaperarXiv: Jailbreaks, Prompt Injection, Poisoning & Defenses

  6. South Korean prosecutors seek three-year sentence in second instance of Chinese national graduate student deepfake sexual crime case

    NewsYonhapEvent

  7. Guatemala launches multi-stakeholder discussions to jointly develop a national artificial intelligence strategy

    Newsprensa.gob.gtAI updates

  8. Irene Liu on why California leads US frontier AI regulation: from the vetoed SB 1047 to SB 53

    NewsThe EU AI Act NewsletterGovernance

  9. An overview of AI safety in California state politics

    NewsLessWrongGovernance

  10. Rumored November 17 global blackout: a mix of distorted facts and fabricated videos

    NewsRTVEEvent

  11. British actors will gain more digital replica rights under an AI protection agreement with producers

    NewsThe Guardian · AIGovernance

  12. Backdoor trigger words are clearly readable in fine-tuned weights

    PaperarXiv: Adversarial Attacks, Backdoors & Poisoning

  13. Conv-Reluplex: A Verification Framework for Convolutional Neural Networks

    PaperarXiv: Adversarial Attacks, Backdoors & Poisoning

  14. Forging a Bearer to take over a LiteLLM MCP session: reproduction and analysis of the OAuth2 passthrough authentication bypass

    NewsXianzhi Community (Aliyun)Event

  15. NATO AI cyber defense uses up to 300B OpenAI tokens per month

    NewsHelp Net Security AIEvent

  16. Google finds AI agents writing and launching exploits without human involvement

    NewsHelp Net Security AIEvent

  17. CuRSE-Bench: A benchmark for probing cultural blind spots in text-to-image safety evaluation

    PaperarXiv: Dangerous Capabilities & Safety Evals

  18. Between Rigor and Reality: How AI Safety Benchmark Developers Understand Benchmark Use and Maintenance

    PaperarXiv: Alignment, Deception & Oversight

  19. Manipulating Bayesian inference and decision-making through data poisoning

    PaperarXiv: 其他 AI 安全论文

  20. Robustness as a Transferable Skill: Cross-Task Adversarial Defense for Vision-Language-Action Policies

    PaperarXiv: 其他 AI 安全论文

  21. The Paradox of Provenance and Privacy: Can Private Cloud Compute (PCC) Prove That Apple Reference Images Are Not Used for Learning?

    PaperarXiv: 其他 AI 安全论文

  22. When tool names are not identity: measuring provider misbinding in MCP-based AI agents

    PaperarXiv: 其他 AI 安全论文

  23. Optimistic verifiers get gamed: reward hacking under exact process oracles

    PaperarXiv: 其他 AI 安全论文

  24. Similar in Safety, Silent in Propagation: Limitations of Text-Similarity Guardrails for AI Agent Security

    PaperarXiv: 其他 AI 安全论文

  25. Evaluating Authorization and Trust Boundaries in Agent Orchestration Frameworks

    PaperarXiv: 其他 AI 安全论文

  26. Jev-as-a-Reward: Validating a Probabilistic Safety Judge for Reinforcement Learning

    PaperarXiv: 其他 AI 安全论文

  27. Experience Rot: Experienced agents fail where novices succeed

    PaperarXiv: 其他 AI 安全论文

  28. Which AI Use Cases Are Worth Piloting? A Risk and Value Framework for Funding, Scoping and Scaling AI

    PaperarXiv: 其他 AI 安全论文

  29. Tight PAC-Bayes Generalization Guarantees for LLM Judges

    PaperarXiv: 其他 AI 安全论文

  30. STA: A Split Temporal Adversarial Method for Evaluating the Robustness of Multimodal Reinforcement Learning Driving Policies

    PaperarXiv: 其他 AI 安全论文

  31. LEAF-OPD: Learning to Unlearn from Privileged Context via On-Policy Distillation

    PaperarXiv: 其他 AI 安全论文

  32. Who Bears the Excess Authority? Attributing AI Agent Behavior to the Delegating Party

    PaperarXiv: 其他 AI 安全论文

  33. Safety-Preserving Fine-Tuning for Medical LLMs via Bi-Level Data Selection

    PaperarXiv: 其他 AI 安全论文

  34. Counterfactual feature steering for identity unlearning with adjacency preservation

    PaperarXiv: 其他 AI 安全论文

  35. Example-based spatial guidance: training-free concept erasure in diffusion models

    PaperarXiv: 其他 AI 安全论文

  36. GPromptShield: Enhancing the Resilience of Graph Prompt Tuning under Adversarial Attacks

    PaperarXiv: 其他 AI 安全论文

  37. SmoothShift: Disrupting Visual Adversarial Attacks against LVLMs at Test Time

    PaperarXiv: 其他 AI 安全论文

  38. Model Immunity Is More Than Condition Number: The Importance of Plateau Regions

    PaperarXiv: 其他 AI 安全论文

  39. Designing Deceptive Model Organisms That Successfully Evade Auditing

    PaperarXiv: 其他 AI 安全论文

  40. Interpretable Federated Learning via Global-Local Attribution Alignment

    PaperarXiv: 其他 AI 安全论文