Skip to content
Oct 13 · Tuesday

All papers

Found 596 papers
Filter6

Attack type

Jailbreaks48Prompt injection43Poisoning and backdoors39Model tampering18Extraction and theft29Detection evasion26Malicious use12

Attack technique

Gradient and representation optimization13Automated search16Encoding and obfuscation7Defense-targeted14Dormant triggers10Selection manipulation3

Model risks

Deception and scheming33Reward hacking26Misalignment and value divergence22Dangerous agent actions25Dangerous capabilities20Inappropriate refusal53Hallucination and fabrication27Bias and discrimination18

Defense type

Model hardening16Inference-time intervention41Detection and filtering17Instruction–data separation9Permissions and approval14Monitoring and auditing34Watermarking and provenance13

Impact

Harmful content85Unauthorized actions52Data and privacy leaks42Output manipulation96Availability and reliability29Oversight failure53Increased misuse capability9

System type

Models and APIs414LLM applications135AI Agent629coding agent127science agent30web agent34Browser and computer use19Multi-agent systems83Embodied systems and robotics27

Attacker access

Black-box18Gray-box6White-box21No-box (offline transfer)3

Attack surface / entry point

User input89Web pages and interfaces4Documents and messages14Tool responses9Tool and skill descriptions25Code repositories and dependencies18Memory writes5Inter-agent messages4Images, audio and video22Training data10Weights and gradients32Distribution channels28Inference services and hardware5

Layer

Prompt layer255Model layer421Application layer175Training and data layer234Infrastructure layer21

System component

MCP and skills26Memory12Guardrails and judges23Monitors43Reasoning chains35Permissions and sandboxing15Fine-tuning and open weights12Vision36Audio15Video5Image generation19Embeddings4Multimodal70
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.
System type：Models and APIsSystem type：coding agentSystem type：science agentSystem type：Embodied systems and roboticsLayer：Model layerLayer：Application layerClear all

Another 213papers 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
  2. OtherarXiv 2610.13912

    RSI-Anything: Unlocking Recursive Self-Improvement with World Models

    Proposes RSI-Anything, which uses a world model to select candidates and retain policy updates.

    arXiv
    2610.13912
    Published
    Layer
    Application layer
    Setting
    coding agent, Browser and computer use
    Evaluated models
    ReasonFlux-Coder-7B, 4B policy, W1
    AbstractRSI-Anything integrates world models into recursive self-improvement through prediction, feedback, and retention checks, and reports gains across multiple domains.

    RSI-Anything is a cross-domain framework: it uses world model predictions to decide which candidates to take external feedback for, then uses the feedback to decide whether the policy, workflow, or predictor is updated, and uses checks to decide which pair is retained. The difficulty to address is the limited interaction budget, and the fact that once the policy changes, the candidate distribution faced by the predictor also changes. The authors argue that improving only one side does not guarantee that the assembled system will be better.

    Paper analysisRead analysis
  4. OtherarXiv 2610.13491

    EviRepair: Evidence-Driven Agents for Reliable Vulnerability Repair

    Proposing EviRepair, an evidence-driven multi-agent framework for repository-level vulnerability repair

    arXiv
    2610.13491
    Published
    Layer
    Application layer
    Setting
    coding agent, Multi-agent systems
    Evaluated models
    DeepSeek-V4.1-Flash, GLM-5.2, Qwen3.8-27B
    AbstractEviRepair uses a reviewed vulnerability model for repository-level repair.

    EviRepair is a repository-level automated vulnerability repair framework for sparse vulnerability reports, using a reviewed vulnerability model to separate diagnosis from patch synthesis.

    Paper analysisRead analysis
  5. OtherarXiv 2610.13479

    A local agentic AI framework generates edge-case scenarios for safety assurance

    Proposes a local dual-model closed loop to generate edge scenarios for autonomous system safety assurance

    arXiv
    2610.13479
    Published
    Layer
    Application layer
    Setting
    LLM applications, Embodied systems and robotics
    Evaluated models
    PPO
    AbstractLocal closed-loop generation of JSON edge cases.

    The authors built a local agentic loop that runs on a laptop to generate edge cases for safety arguments of autonomous systems, rather than releasing a new public test set.

    Paper analysisRead analysis
  7. OtherarXiv 2610.13372

    Toward Safe World Models: Risks, Defenses, and Evaluation

    A survey of risks, defenses, and evaluation in the six-stage lifecycle of world models

    arXiv
    2610.13372
    Published
    Layer
    Application layer
    Setting
    Embodied systems and robotics
    AbstractSafety is written as a property of the prediction-decision closed loop; claims must be linked to observed consequences and versions.

    World model safety is organized by the nature of the prediction-decision loop, not by isolated prediction scores.

    Paper analysisRead analysis
  10. 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
  11. InterpretabilityarXiv 2610.12989

    GPT-2-Small, Gemma-2-2B, and Llama-3.1-8B have interpretable narrative mechanisms

    Using NVC to locate and ablate narrative-related components in language models

    arXiv
    2610.12989
    Published
    Layer
    Model layer
    Setting
    Models and APIs
    Evaluated models
    GPT-2-Small-124M, Gemma-2-2B, Llama-3.1-8B
    AbstractNeurons and SAE features related to narrativity across three models, when ablated, changed some narrative-related behaviors.

    The authors connect a three-dimensional definition from narratology to mechanistic interventions, examining whether components related to narrativity in pretrained language models participate in prediction, generation, and downstream tasks.

    Paper analysisRead analysis
  12. InterpretabilityarXiv 2610.12842

    Neural networks are more modular than single neurons suggest

    Using a three-level nested decomposition to test whether mixed-selectivity networks can be functionally separated on subspaces

    arXiv
    2610.12842
    Published
    Layer
    Model layer
    Setting
    Models and APIs
    Evaluated models
    Pythia 410M, Pythia 160M, GPT-2 XL and others18
    AbstractUnit non-specialization and subspace separability can coexist.

    The authors propose a nested definition of modularity to separate "neurons not being specialized" from "computation not being separable."

    Paper analysisRead analysis
  13. 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
  14. 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
  15. Evaluation and benchmarksarXiv 2610.12569

    SWE-Duel: Toward a Self-Expanding Arena for Coding Agents

    Proposes SWE-Duel, which lets coding agents mutually evaluate, create problems, and blindly review fixes during evaluation.

    arXiv
    2610.12569
    Published
    Layer
    Application layer
    Setting
    coding agent
    Evaluated models
    GPT-5.5, Claude Opus 4.8, Kimi K2.7 Code and others11
    AbstractThe mutual evaluation arena creates questions during evaluation.

    SWE-Duel is a coding agent arena that generates tasks and judges only at evaluation time, to avoid leakage and saturation of static benchmarks.

    Paper analysisRead analysis
  16. InterpretabilityarXiv 2610.12410

    The paper proposes an alignment generalization prediction task, using activation representations to predict the cross-context effects of fine-tuned values.

    Comparing the predictive power of value representations for alignment generalization after single-value fine-tuning

    arXiv
    2610.12410
    Published
    Layer
    Model layer
    Setting
    Models and APIs
    Evaluated models
    Olmo-3-7B, Olmo-3-32B, Qwen-3-8B-Base and others4
    AbstractActivation representations can predict cross-value generalization before training and relate to robustness across multiple value objectives.

    The authors define "how single-value fine-tuning changes compliance on untrained values" as a task predictable before training, and use value representations to do this.

    Paper analysisRead analysis
  17. Analysis and theoryarXiv 2610.12409

    The researchers conduct a psychometric review of HELM Safety and HarmBench, questioning whether their scores can represent a single attribute

    Auditing whether HarmBench scores can represent a single harmful refusal attribute

    arXiv
    2610.12409
    Published
    Layer
    Model layer
    Setting
    Models and APIs
    AbstractThe authors claim HarmBench does not correspond to a single harmful refusal, and behaviors should be separated when comparing models.

    This is a construct validity audit of HarmBench in HELM Safety, examining whether harmful refusal can be read as a single attribute when interpreting scores.

    Paper analysisRead analysis
  19. 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
  20. OtherarXiv 2610.12235

    The paper proposes using language models as world models for AI research to predict experimental results.

    Propose studying world models, using language models to predict intervention gains

    arXiv
    2610.12235
    Published
    Layer
    Model layer
    Setting
    Models and APIs
    Evaluated models
    Claude Opus 5, Grok-4.6, GPT-6 Astra and others13
    AbstractThe authors claim experimental records can be reused inside and outside the environment and improve experiment selection under fixed budgets.

    The authors use language models as research world models, predicting the median intervention gain before execution, to compare candidate experiments under a limited budget. The gap to address is that research agents propose faster than they execute. The same type of change can have different effects across reference models, recipes, and budgets, so experimental knowledge that can continue to be used outside the original environment is needed.

    Paper analysisRead analysis
Load more