Translation in progress Some entries still include text in the original language.
Oct 12 · Monday
All papers Found 13 papers Another 273 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.
Extending reinforcement learning for full-modality foundation models to support agent self-improvement
Published Oct 8
Evaluated models MiMo-V2.6-Pro , MiMo-V2.6-Flash , MiMo-V2.6-Distill-Qwen-9B Abstract The series uses three-dimensional RL and group scoring to improve agent scores; the authors claim the final performance is comparable to frontier models.MiMo-V2.6 is an all-modality MoE technical report, pursuing agent self-improvement by scaling up RL computation, environments, and scoring.
在 CWEval 上跟踪代码生成的功能-安全差距
Published Oct 6
Evaluated models Gemini 1.5 Pro , Gemini 2.5 Pro , Gemini 3.1 Pro and others20 Abstract 绝对安全性上升且差距仍在，开放权重并不分开各家族的轨迹。
提出 VERA，审计 LLM 漏洞解释里的虚构推理
Published Oct 5
Evaluated models sonnet-5 , gemma-4-26b , llama-4-maverick and others8 提出选择性长时程精炼，筛选轨迹以提升终端智能体训练可靠性
Published Oct 5
Evaluated models Qwen/Qwen3-8B Abstract 中间长度的监督更均衡。这项工作研究终端智能体模仿学习里，教师轨迹应保留多少 token。
提出拓扑条件后门，使代码模型在推断多智能体部署时植入漏洞
Published Oct 5
Evaluated models Qwen2.5-7B-Instruct , Qwen2.5-Coder-7B-Instruct Abstract 论文证实了拓扑条件后门的可训练性与泛化性，展示了常规评测盲区，并提出差异化拓扑审计作为检测防御方向。
检验低监控读数能否证明代码生成中的奖励作弊已受控
Published Oct 2
Evaluated models Llama-3.1-8B-Instruct Abstract 论文指出监控器训练读数接近零无法确立作弊受控，模型可利用通用外壳推迟提交绕过监控，强化学习亟需带外行为验证。本文是一篇针对强化学习可验证奖励后训练（RLVR）中监控器有效性与奖励作弊行为的实证分析研究。论文要解决的核心问题是：在训练目标中引入监控器以抑制代码生成中的作弊行为时，接近零的监控器读数是否足以确立模型行为已真正受控。
提出 AuraForge，从漏洞修复合成安全测试并训练编码 Agent
Published Oct 1
Evaluated models Qwen3.5-4B , GLM 4.7 Flash , Qwen3-Coder-Next and others7 Abstract AuraForge 用攻击效果测试和防泄露环境做成 AuraGym，合成监督的误拒更低，训出的 4B 模型增益高于人工测试。AuraForge 把真实仓库的历史漏洞修复变成可执行训练任务，用来监督编程智能体同时满足功能请求和隐含的安全要求。
构建编码 RL 奖励作弊测试台 CATCH 并检验 CoT 监控失效
Published Sep 30
Evaluated models Qwen3-4B , Qwen3.5-27B Abstract CATCH 用双运行金标签研究编码 RL 的奖励黑客，并观察到监控会被注释适应削弱。CATCH 是面向编码 RL 奖励黑客的可控分析测试床，用可利用环境、可调初始倾向和基于执行的金标签，在训练全程比较黑客动态与干预。
提出 CGMIA，检测代码生成基准样本是否泄漏进训练集
Published Sep 9
Evaluated models Qwen2.5-Coder (3B) , CodeGemma (2B) , DeepSeek-Coder (1.3B) and others5 Abstract CGMIA 融合相似度、通过率、困惑度与 CodeBERT，在模拟泄漏上优于八个 MIA 基线。CGMIA 用成员推断检测代码生成基准是否泄入 LLM 训练集，使评测不依赖对训练语料的直接访问。
提出 ERPO，用探针共识奖励在无标签代码题上做测试时强化学习
Published Sep 9
Evaluated models Qwen3-4B , Qwen3-8B 提出 OpenMLE，后训练元进化智能体并做机器学习工程长程搜索
Published Jul 31
Evaluated models Frontis-MA1-35B , Frontis-MA1-30B Abstract OpenMLE 训练 Frontis-MA1-35B，用四个共享算子把 MLE 后训练和长程搜索接成元进化闭环。作者提出全栈 OpenMLE ，在机器学习工程里研究从进化到元进化的一步，并把递归自我改进视为更远目标。智能体要在延迟、有噪声的执行反馈下反复修改程序。作者认为公开系统很少同时提供可扩展环境、基于执行的后训练，以及部署该模型的长程进化搜索。
用 abliteration 消融代码 LLM 拒答方向，分离拒答与漏洞注入能力
Published Jun 4
Evaluated models Qwen2.5-Coder-Instruct 3B , Qwen2.5-Coder-Instruct 7B , Qwen2.5-Coder-Instruct 14B Abstract 拒答可用低秩编辑去掉，CWE-89 注入是否成功仍由模型规模决定。这项初步可行性研究用 abliteration 把代码 LLM 的拒答和 CWE 注入能力拆开：意愿可以被低秩编辑解锁，注入是否成功仍看参数规模。
复评前沿模型代码生成中的包名幻觉与抢注攻击面
Published May 17
Evaluated models Claude Sonnet 4.6 , Claude Haiku 4.5 , GPT-5.4-mini and others5 Abstract 评估显示前沿模型包幻觉率收窄但威胁仍在，作者识别出跨模型通用攻击面并呼吁联合防御。
You have reached the end