Translation in progress Some entries still include text in the original language.
Oct 12 · Monday
All papers Found 91 papers Another 265 papers do not have category tags yet and are not included in filtered results.
“
About this collection Collection in progress We are building a large AI safety paper collection to help researchers find work related to their interests. Papers are collected automatically andfiltered only for relevance to AI safety , without judging quality, novelty or reading value. The default “Annotated first” shows papers with tags or published paper analyses first. Each group is ordered by original publication date, newest first, with undated papers last. “Newest first” uses publication dates only. This order is not a quality ranking.
One-line summaries, tags and paper analyses are prepared by AI from the papers. Language models may have systematic biases and produce hallucinations.We plan to carry out human annotation and review to reduce errors in this content.
Proposes RSI-Anything, which uses a world model to select candidates and retain policy updates.
Published Oct 10
Evaluated models ReasonFlux-Coder-7B , 4B policy , W1 Abstract RSI-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.
Proposes RATED, which reranks agent tool retrieval using execution feedback
Published Oct 10
Evaluated models DRAGON Proposes a local dual-model closed loop to generate edge scenarios for autonomous system safety assurance
Published Oct 9
Evaluated models PPO Abstract Local 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.
Proposes EndoRSI, recursively improving endoscopic robot policies with physical evidence
Published Oct 9
Evaluated models Qwen3-VL-4B A survey of risks, defenses, and evaluation in the six-stage lifecycle of world models
Published Oct 9 Abstract Safety 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.
Propose COAP, which uses stateful code to control a robot at test time without a model
Published Oct 9 Proposes recursive self-improvement based on rulebooks, enabling a frozen LLM to learn an explicit world model.
Published Oct 8
Evaluated models Claude Opus 5 Abstract A frozen LLM uses a revisable rulebook and an executable model to learn the environment, and reaches the RHAE upper bound on ARC-AGI-3.
提出 GROB，用只读公开痕迹保存候选智能体线索并检验归因边界
Published Oct 8 Abstract 作者称 Census 是最强例。GROB 是一套多角色架构，在没有特权遥测、也不预先指定目标时，从公开网页痕迹里保存候选自主智能体活动，供日后解析。
提出 GENUI-Harness，用生成界面收集未决输入并完成数据库服务任务
Published Oct 8
Evaluated models GENUI-4B , Qwen3.5-4B , Qwen3.6-27B and others10 Abstract 数据感知界面加奖励审计。GENUI-Harness 把数据库服务任务的交互，从纯文本改成按当前请求和检索记录生成的界面。
提出 System Switch，让快决策模型仅在门控打开时询问推理模型
Published Oct 7
Evaluated models Laya, typed-decisions , Laya, base , Kev 9B and others12 Abstract 作者称快模型应按敏感度选。System Switch 在不暂停的 Doom 里，让快的 typed-decision 模型做每一次决策，只在门控打开时把控制交给视觉推理模型。
用混合生成式多智能体仿真考察基准留出对评分与满意度差距的影响
Published Oct 7 发布开源个人智能体 nanoMuse，经 Sentinel 审查设备屏幕动作
Published Oct 7 提出 Transect，将长程 Agent 评测转录对齐到可回查的共享时间线
Published Oct 6
Evaluated models Claude-Opus-4.7 Abstract Transect 把长转录收成可回查报告，案例中文稿与自审占主要 token。Transect 是基于 Inspect Scout 的转录分析包，用来在长程智能体评估里保住可回查的过程账户。
定义任务充分收缩，认证用 GC 替换全信源后速率–遗憾曲线不变
Published Oct 6 Abstract 任务可在优化前定出压缩信源。任务充分收缩 处理机器信息接口上的信源选择：已声明任务有时可以在编码器、码本、速率、失真目标或优化器选定之前，定出一个更小的信源。
提出 DAEDALUS，用自生成任务自举 Agent 记忆
Published Oct 6
Evaluated models GPT-5.4-mini , GPT-5.6 Luna , Qwen3.6-35B-A3B and others11 提出 AI Agent 主张的可审计性框架与裁决规则
Published Oct 6 Abstract 主张须事先固定四要素。作者给出把关于已部署智能体的主张写成可核查陈述的条件，而不是检测器或攻击方法。
提出 NetAgent，编排工具完成多任务网络流量分析
Published Oct 6 Abstract NetAgent 编排工具做多任务流量分析。NetAgent 是一个多任务网络流量分析智能体框架。作者要让分析员用自然语言把分类、隧道检测和主机研判串成可重规划的工作流，而不为每个任务单独训练一个模型。
提出 HERA，协同演化 harness 与环境以提升 Agent 拒答准确性
Published Oct 5
Evaluated models GPT-5.6-Luna , GPT-5.6-Terra , GPT-5.6-Sol and others20 Abstract HERA 用失败协同演化 harness 与环境，并在迁移评测中提高 Abstain。HERA 在模型权重固定时，让 harness 与可执行环境按失败协同演化，以改进 agentic abstention。
蒸馏私有健康记录为疾病特异性打分器，供外部智能体发现血液标志物
Published Oct 4 Abstract 私有队列被蒸馏成 GAT 打分器。把私有 CBC 证据蒸馏成可发布的 GAT 打分器，使前沿智能体在不读取患者行时给生物标志物表达式排序。 。
提出 WebFovea，加固视觉网页智能体的解析、执行、反馈与观察往返
Published Oct 2
Evaluated models claude-opus-4-6