AI 论文与研究

OmniEdu 发布面向 K-12 学习与教学的开源基础模型家族，覆盖学科能力、课程理解、诊断推理与教学行动四项能力，微调出 4B、9B、27B 三个规模。其指令微调语料含 69，999 条样本、15.96M 监督回答 token，其中教育专用样本 60，951 条。

RoboDawn 通过离散平移、旋转和夹爪命令，让智能体式 VLM 以闭环方式控制机器人，并借助上下文学习用少量演示完成界面使用与任务策略落地。在 RoboTwin 2.0 C2R 上，成功率从零样本 53.2% 提升到单样本 73.6%，超过基线 π0.5 的 46.0%；在 RoboDojo 上从 35.67% 提升到 47.17%，且无需针对任务做机器人训练。

RewardVerse 提出基于评分标准（rubric）的视频奖励框架，在评估查询与打分器之间引入动态 rubric 作为中间表示，先显式生成评估标准再进行引导式打分，以缓解直接映射单一分数导致的 scalar drift。

针对全流程 FP8 强化学习仍存在训练不稳定、中期熵激增和输出乱码的问题，研究揭示其根源是 FP8 量化噪声叠加扭曲重要性比值，使负优势 token 被错误推出信任域并梯度归零。

Spatial-Interactor 框架通过交互轨迹训练 VLM 建模物理世界状态转换，采用 L1 被动世界状态、L2 主动自身状态、L3 长时程交互轨迹三级课程，并构建了 LSI-108K 数据集。

UltraTex 是一个面向高分辨率多视角扩散 3D 纹理生成的端到端框架，可将工作分辨率从 512/768 提升至 2048。

论文提出自组织智能体团队 SAT，让固定团队的智能体从以往协作中学习角色分工、对话阶段和信息流策略，实现协同计算。仅用 15 道数学和 25 道研究生级知识题学到的策略可原样迁移到未见基准，五个数学与物理基准平均准确率 66.7%，高于最强成员的 48.8% 和完美路由器的 59.0%，在 AIME 2026 上超出该路由器 13.4 分。

谁会成为第一个在RL或评测过程中意外黑进一家公司的学术实验室。那肯定比最佳论文奖或奖学金更有价值。

AMD 警报🚨🚨：在 Kimi K3 2.8T 上，AMD 每吉瓦的利润率甚至优于 B300、GB200 NVL72 和 B200。这得益于 AMD 更低的 TCO 和更好的性能，Mi355X 的利润率达到 53.6%，而 GB200 NVL72 为 44.3%。这来自我们的开源 Agentic Inference 基准测试 AgentX。 感谢 @roaner 和 @EmadBarsoumPi 两位 10x 工程师优化了这个 2.8T 模型！

Ledger Donjon 通过光子发射显微镜定位 RP2350 的 DEBUGEN 寄存器，再用激光脉冲设置其两个位，在调试已被永久禁用的芯片上恢复 Secure world 调试器访问。

Chamath 撰文指出，开源权重模型与前沿闭源模型的能力差距仅约 4 个月，因此不必赢下所有 benchmark，只需让闭源模型的溢价变得可谈判。对许多日常企业负载而言，更低的成本、更深的定制、基础设施控制权以及把专有数据和工作流留在公司内部，可能比略微落后于前沿更重要。市场将走向混合格局，而非赢家通吃。

JEPA-Anything 在不同世界间学习预测模型 论文：https://huggingface.co/papers/2609.20800

编码智能体Harness设计的实证研究 论文：https://huggingface.co/papers/2609.20804

Google Research 推出 MilleMiglia，一个面向中程物流（middle-mile）的真实实例生成器。中程物流不同于首程和末程，单件货物需经多辆车辆接力运输，在多个配送中心间中转、分拣并合并，且必须在指定时间窗口内赶上下一班卡车，否则将滞留至下一周期造成显著延误。

Professional graphic design is a long-horizon agentic task in which structured, editable artifacts emerge from many interdependent actions, yet outcomes admit no reliable programmatic oracle. We introduce a continual adaptation framework i…

Occupational accident narratives contain valuable information about work situations, unfavourable conditions, accident events, and their consequences. Automatically structuring these narratives can facilitate large-scale accident analysis …

Training capable coding agents via reinforcement learning (RL) requires diverse tasks with reliable verifiers. Open-source codebases offer a rich source of such tasks, while existing methods typically rely on development artifacts such as …

Users increasingly delegate work to autonomous AI agents, yet evaluations typically measure task completion rather than the values users prioritize. Using Value Sensitive Design, we analyzed, with LLM assistance, 73,093 first-person Reddit…

Advances in large-scale neural recording have made it possible to collect data across many animals and distributed brain regions, raising the question of whether this scale can be exploited to learn general-purpose neural representations t…

Multi-hop retrieval failures are not uniformly distributed across queries: they cluster in structurally predictable subpopulations. We prove two results formalizing this structure. First (CWAR Reducibility): confident-failure reduction is …

A desirable property of a world model is the ability to learn continually across tasks, adapting to new environments without forgetting what the agent has already learnt. In particular, the ability to retain and reuse knowledge obtained fr…

Graph Convolutional Networks (GCNs) are highly sensitive to label noise, since corrupted supervision can propagate through the graph and degrade learned node representations. This work proposes PCC+GCN, a hybrid framework that uses Particl…

A selective predictor acts as a safety gate: it returns an output only when the prediction appears sufficiently trustworthy. Deployments increasingly require this reliability to be certified at a target precision for every reporting unit o…

Memory systems for large language models have focused predominantly on efficient retrieval, whereas the decision of whether retrieved memories should be trusted has received comparatively little attention. When the memory store contains co…

Reinforcement learning is increasingly used to align image generators with reward signals, and Flow-GRPO recently extended this paradigm to flow-matching models by treating the denoising sampler as a stochastic policy that can be optimized…

We need studies on conversational AI (CAI) at scale to understand human behavior and shape CAI design. However, fragmented reporting of systems and study configurations hinders replication, extension, and knowledge accumulation. We present…

We introduce QuranicMMLU, a benchmark for evaluating generative AI on Quranic Arabic across multiple dimensions of linguistic complexity. Existing Quranic benchmarks center on general question answering and semantic retrieval, without prob…

Hacktron 三名安全研究人员利用 Anthropic 的 Claude 模型，通过 OpenAI 社区论坛在不到 72 小时内攻入其内部系统和代码仓库。

OpenAI 的 Noam Brown 回应其"气隙隔离不足以阻止失控 AI"的言论，称该温度传感器通信例子纯属学术假设，本意是说明绝对隔离保证极难实现，因此需要多层防御。他强调该例子讲的是本应完全隔离的智能体之间的协调，协调只需极少信息比特。他还提到 HF 事件的教训是过度信任沙箱隔离、缺乏独立防护，设计安全协议时应高估而非低估 AI。

Every multiparameter persistence vectorization we know of carries a one-sided Lipschitz upper bound and nothing below it: without a lower gauge there is no sense in which the features are faithful, and no per-prediction guarantee can be bu…

Vision-language models (VLMs) rely on storing and transferring appropriate information across their sub-components. Verifying that the VLMs exhibit desired behaviours, while avoiding harmful ones, is central to their reliable deployment. Y…

Gating the value pathway of attention reportedly improves language model pretraining, and prior studies disagree on why. We argue and provide experimental evidence that such gates supply two different things that softmax attention lacks: a…

Computer-use agents (CUAs) have advanced along two separate lines: graphical interaction and software development through code and the command line. Real digital work requires both, interleaved rather than stacked end to end. We study hybr…

LLM agents in social simulation revise their opinions implicitly, in context: how open an agent is to persuasion can neither be specified nor verified, and collective outcomes inherit the model's training prior. We introduce Bayesian Chron…

Large language models can hold knowledge they do not report. A model may sandbag on a capability evaluation, or answer against what it internally knows, and its outputs alone cannot tell whether it is hiding an answer or simply does not ha…

The prediction of critical heat flux (CHF), a key safety-related quantity in nuclear thermal hydraulics, remains an important challenge due to its direct relationship with fuel performance and reactor safety. Recent studies have demonstrat…

腾讯混元研究团队从第一性原理分析 LLM 强化学习中的 batch size scaling，提出以 time-to-target 为判据：只有当吞吐增益 r_q 超过样本代价 r_N 时，更大的 batch 才能缩短训练时间。

Empirical evaluation is central to recommender-systems (RecSys) research, but turning experimental designs into executable code remains a manual and error-prone task. We present AutoRecLab, a Python-based autonomous RecSys lab that automat…

Nearly 90% of drugs entering clinical development ultimately fail, despite billions of dollars in investment. Pharmaceutical companies therefore rely on clinical development planning (CDP) and probability of technical and regulatory succes…

Large language models (LLMs) have achieved state-of-the-art performance across a wide range of tasks, motivating two important aspects of deployment: inference efficiency and output provenance, which can be tackled by speculative sampling …