News | Embodied Intelligence Observer

Aug 27 Thu25 articles

Research
AI’s memory crunch is coming for Android apps
Google is setting new memory-use limits for Android apps as AI data centers contribute to hardware shortages that could leave lower-cost phones with less memory.
22:27TechCrunch AIOriginal ↗
Policy / Standards Expos / Launches Research
百度搭子宣布一揽子升级，提出“交付即惊艳”专业办公标准
8月27日，“搭子进行时——百度搭子产品发布会”在上海举行。百度搭子围绕“交付即惊艳”宣布个人版、企业版、专业套件及工作台能力一揽子升级，重点面向自媒体、金融等专业场景，覆盖选题策划、资料研究、内容制作、财报分析、行业研究和报告输出等任务，并形成可用于发布和汇报的图文、脚本与分析报告。此前发布的AI产品榜显示，百度搭子月活674.30万，环比增长1063.79%，位列AI办公智能体增速第一。百度集团执行副总裁、百度智能云事业群总裁沈抖在现场表示，百度搭子今年3月发布以来，5个月内已完成150次迭代，过去一个月用户增长近9倍。用户规模增长快，说明产品命中了用户需求，但产品体验是否做到极致、交付结果是否做到惊艳，决定用户是否愿意留下。百度搭子要用更好的体验和更稳定的交付，让智能体真正成为个人和企业值得依赖的生产力伙伴。资料分散、工具频繁切换、生成结果仍需大量修改，是当前AI办公的主要问题。百度搭子此次提出“交付即惊艳”，将办公智能体的价值落到最终成果：执行过程专业，结果内容扎实、表达出彩、拿来即用，同时允许用户随时修改、接管和复用。针对不同任务挤在同一个对话框、结果难以继续加工的问题…
19:27雷峰网Original ↗
VLA / Models Research
像素级对标？智谱、Kimi 底层参数「撞衫」背后，藏着线性注意力的黄金窗口
国产大模型内卷到算子底层。作者丨高允毅编辑丨岑峰昨晚，智谱 AI 认领了最近一周在 OpenRouter 盲测榜上风头最盛的“牛来”模型，正式命名GLM-5.3-Flash。这款上线首日就登顶 OpenRouter 调用量榜单，刷新全网长文本性价比斩杀线的国产模型，在底层架构上展现出“集百家之长”的特质。它采用了混合架构，将 DeepSeek 的稀疏注意力机制（NSA）与 Kimi 赖以成名的线性注意力机制结合，再叠加DeepSeek的流形约束超连接（mHC）技术，可以大幅降本提效。今天早上，月之暗面核心研究员、“注意力残差”论文作者陈广宇，仔细研究了 GLM-5.3-Flash 的模型配置文件后，贴出一张 Kimi K3 和 GLM-5.3-Flash 的底层代码对比图。他发现两者不仅都采用了 KDA 线性注意力，排布逻辑高度重合，连核心参数“遗忘门下界”（gate_lower_bound）同样是 - 5。从 Kimi K3 公开技术报告到智谱上新，还不到一个月。有人感叹：哪怕是参考了 Kimi 的公开参数，短短 20 天内就能推出这么强的模型，也太惊人了。 …
19:01雷峰网Original ↗
Research
Submit Your Questions: The Great Data Center Backlash
You have questions about data centers, and WIRED has answers. Join our livestream on September 10 and our panel of experts will tell you everything you need to know.
19:00Wired AIOriginal ↗
VLA / Models Funding Research
亚马逊 Mechanical Turk 将于 9 月 30 日关闭
亚马逊宣布其众包平台 Mechanical Turk 将于 9 月 30 日关闭。亚马逊上个月才宣布将于 7 月 30 日起停止接受新用户。当时亚马逊表示该决定是在“慎重考虑”后做出的，“现有用户可以继续正常使用该服务。AWS 将继续投资改进 Mechanical Turk 的安全性和可用性，但我们不打算推出新功能。如今它正式给 Mechanical Turk 画上了句号。亚马逊是从 2018 年起将 Mechanical Turk 变成训练神经网络的标注数据服务。但讽刺的是 2023 年的研究发现，该平台 33% 到 46% 的众包工作者使用大模型去完成任务，引发了对标注数据的可靠性以及是否真的需要人类参与的质疑。由于大量的机器人和欺骗行为，研究人员已经放弃了该平台，它的关闭只是时间问题。
18:48奇客SolidotOriginal ↗
VLA / Models Commercialization Open Source
大模型推理加速全链路：内存管理、编译优化、量化与并行策略
大模型推理成本已成为AI产业落地的核心瓶颈。随着模型参数规模向万亿级迈进、多模态智能体应用场景爆发，推理引擎需要应对算子实现、内存管理、量化压缩、异构调度与并行策略等多维度的技术挑战。本文整理自清华大学助理研究员金煜阳博士在 QCon 全球软件开发大会 2026 北京站的分享《大模型推理加速全链路：内存管理、编译优化、量化与并行策略》。金煜阳系统介绍了其所在实验室在大模型推理加速全链路方面的探索，覆盖从底层算子优化到上层并行策略的完整技术栈。本文将逐一展开算子级性能调优中的细粒度图层优化方法、面向异构模型结构的KV Cache内存管理、编译与运行时协同的混合精度量化、基于负载特性的CPU-GPU异构调度，以及面向动态请求场景的自适应并行策略，最后介绍开源推理引擎赤兔在国产芯片生态中的实践成果。以下是演讲实录（经 InfoQ 进行不改变原意的编辑整理）。背景与挑战人工智能的市场规模正在以超出预期的速度扩张。去年来自Precedence Research和艾瑞咨询的数据预测，到2030年全球AI市场规模将达到11万亿。但今年年初OpenAI的持续爆发，以及智能体应用的集中涌现，让这…
17:47InfoQ 中文Original ↗
World Models Funding Autonomous Driving
ECCV 2026 倒计时 12 天：LeCun 压轴、中国 71M 模型、CMU 一键修 3D，顶会三大爆点抢先看
今天，距离 ECCV 2026 在瑞典马尔默开幕只剩 12 天，会前预热迎来最重磅的一天：官方完整议程首次揭晓，图灵奖得主 Yann LeCun 携手自动驾驶与世界模型领域的两位顶尖学者压轴 Keynote；与此同时，中国团队带来仅 71M 参数的实时人像“全能模型”，CMU 则拿出能一键修复 3D 渲染瑕疵的通用视频模型。这些进展不仅勾勒出本届大会“世界模型 + 自动驾驶 + 轻量实时生成”的主线，也为关注 AI 视觉落地的从业者、研究者与投资人提供了清晰的观察坐标与持续跟踪清单。会议情况欧洲计算机视觉会议（ECCV）与 CVPR、ICCV 并称计算机视觉三大顶会，偶数年举办。本届将于 9 月上旬在瑞典马尔默举行，整体会期与关键信息如下：名称 ECCV 2026（第 19 届欧洲计算机视觉会议）时间 2026 年 9 月 8–12 日（Workshops/Tutorials 9/8–9；主会与 Expo 9/10–12）地点瑞典马尔默，Malmö Arena 与 Malmömässan 会展中心官网 https://eccv.ecva.net/ 时区马尔默 CEST…
16:13雷峰网Original ↗
Funding Policy / Standards Research
OceanBase以20.2%市场份额居首，中国分布式数据库迈入规模化竞争
中国分布式数据库市场正在从“技术可行”走向“规模化竞争”。近日，全球增长咨询公司沙利文发布《2026年中国分布式事务型数据库市场研究报告》，首次从技术标准层面对分布式数据库进行明确界定。报告显示，2025年下半年至2026年上半年，中国分布式事务型数据库市场规模约59.5亿元，OceanBase以20.2%的市场份额位居第一。与此同时，中国分布式数据库市场已从互联网及支付场景的内部技术验证，逐步进入金融核心系统突破、通信及关键行业规模应用阶段。伴随数据库技术持续成熟、生态体系不断完善，以及企业对核心业务系统可靠性要求不断提高，分布式数据库正逐步从新型数据库产品，发展为企业数据库体系的重要组成部分。分布式数据库进入规模化竞争期此次研究首先厘清了分布式数据库的市场边界。报告进一步从技术架构层面对分布式数据库作出明确界定：采用存储与计算均可横向扩展的分布式架构、具备跨节点事务一致性、数据强一致性及高可用能力，能够实现 RPO=0 的数据可靠性保障。按照上述定义，2025H2—2026H1期间，OceanBase以20.2%的市场份额位居第一，阿里云、腾讯云和华为云分别以17.9%、…
14:59雷峰网Original ↗
Open Source Research
亚马逊收购开源数据库 DuckDB 开发团队
亚马逊同意收购开源数据库 DuckDB 开发团队 DuckLabs。DuckLabs 员工将加入亚马逊 AWS，其中包括联合创始人 Hannes Muhleisen 和 Mark Raasveldt，他们将继续领导团队和制定项目的技术方向，员工也将继续在阿姆斯特丹办公。DuckDB 项目将继续维持现有的开源状态，使用 MIT 许可证，由独立基金会管理。收购 DuckLabs 被认为有助于将亚马逊的云存储服务 S3 转型为客户分析数据而非仅仅存储数据的平台。
14:47奇客SolidotOriginal ↗
Commercialization Research
从数字化到AI：企业数智跃迁的路径、实践与挑战 | 2026 ITValue Summit前瞻：AI系统落地系列直播②
对企业来说，下一阶段更值得关注的，是如何找到真正有价值的场景，在可控范围内不断试验，并让技术、业务和人的角色在实践中重新磨合。
12:08钛媒体Original ↗
Research
Point2Pose: Occlusion-Recovering 6D Pose Tracking and 3D Reconstruction for Multiple Unknown Objects Via 2D Point Trackers
arXiv:2604.10415v2 Announce Type: replace-cross Abstract: We present Point2Pose, a model-free method for causal 6D pose tracking of multiple rigid objects from monocular RGB-D video. Initialized only from sparse image points on the objects, our approach tracks multiple unseen objects without requiring object CAD models or category priors. Point2Pose leverages a 2D point tracker to obtain long-range correspondences, enabling instant recovery after complete occlusion. Simultaneously, the system in…
12:00arXiv cs.ROOriginal ↗
VLA / Models Datasets Research
Regularized Latent Dynamics Prediction is a Strong Baseline For Behavioral Foundation Models
arXiv:2603.15857v2 Announce Type: replace-cross Abstract: Behavioral Foundation Models (BFMs) produce agents with the capability to adapt to any unknown reward or task. These methods, however, are only able to produce near-optimal policies for the reward functions that are in the span of some pre-existing state features, making the choice of state features crucial to the expressivity of the BFM. As a result, BFMs are trained using a variety of complex objectives and require sufficient dataset co…
12:00arXiv cs.ROOriginal ↗
Research
Trojan Attacks on Neural Network Controllers for Robotic Systems
arXiv:2602.05121v3 Announce Type: replace-cross Abstract: Neural network controllers are increasingly deployed in robotic systems for tasks such as trajectory tracking and pose stabilization. However, their reliance on potentially untrusted training pipelines or supply chains introduces significant security vulnerabilities. This paper investigates backdoor (Trojan) attacks against neural controllers, using a differential-drive mobile robot platform as a case study. In particular, assuming that t…
12:00arXiv cs.ROOriginal ↗
Research
Three-Way Open-Set Detection for Robust Autonomous Navigation
arXiv:2511.15343v2 Announce Type: replace-cross Abstract: Autonomous navigation in complex scenes requires reliable perception across scenarios that the model did not encounter during its training. Along its route, an autonomous framework encounters objects it was trained to recognize, obstacles it has never seen, and background structures that resemble objects. Each of the three must be handled differently. To tackle this, existing open-set and out-of-distribution detectors discard low-confiden…
12:00arXiv cs.ROOriginal ↗
Research Safety / Ethics
Model-Agnostic Open-Set Air-to-Air Visual Object Detection for Reliable UAV Perception
arXiv:2509.09297v2 Announce Type: replace-cross Abstract: Open-set detection is crucial for robust UAV autonomy in air-to-air object detection under real-world conditions. Traditional closed-set detectors degrade significantly under domain shifts and flight data corruption, posing risks to safety-critical applications. We propose a novel, model-agnostic open-set detection framework designed specifically for embedding-based detectors. The method explicitly handles unknown object rejection while m…
12:00arXiv cs.ROOriginal ↗
Reinforcement Learning Research
Learning to Act While Waiting: RL Finetuning of Generalist Robot Policies Under Inference Latency
arXiv:2608.23831v2 Announce Type: replace Abstract: While reinforcement learning (RL) allows generalist robot policies to continually improve during deployment, the large model size of modern generalist policies, such as VLAs, poses a fundamental obstacle to effective RL improvement. In particular, their severe inference latency---which can lead to pauses or jerky movements---can alter the effective environment dynamics and, if not correctly accounted for, break the Markov assumption that RL rel…
12:00arXiv cs.ROOriginal ↗
VLA / Models Research
Meta-Ctrl: Guaranteed Plan Generation by Decoupling Syntactic and Semantic Constraints
arXiv:2608.22149v2 Announce Type: replace Abstract: LLMs generate fluent plans for robots but routinely violate the syntactic and se8mantic constraints they must satisfy to execute, and existing remedies trade formal guarantees against plan quality: soft methods (affordance scoring, grounded decoding) give no guarantee, while symbolic planners (LLM+P) discard the LM's commonsense. We propose \textbf{Meta-Ctrl}, a constrained-decoding framework that guarantees the encoded constraints while preser…
12:00arXiv cs.ROOriginal ↗
Research
Min-Max Regret Task Allocation and Planning of Heterogeneous Multi-Robot System in Partially Known Environments
arXiv:2607.13403v2 Announce Type: replace Abstract: Efficient task allocation for large-scale Heterogeneous Multi-Robot Systems (HMRS) is critical, yet dealing with complex temporal logic tasks in partially known environment (PKE) remains a computational bottleneck. Existing approaches often struggle to balance exploring uncertain regions and exploiting known resources, while also suffering from exponential computational complexity. To address these issues, this paper presents a robust planning …
12:00arXiv cs.ROOriginal ↗
Research
RoboMME-Interference: Benchmarking Robot Memory Under Interference
arXiv:2606.22338v3 Announce Type: replace Abstract: Robots deployed in realistic settings will accumulate experience across many sessions and tasks over their deployment. The robot's tasks may often require it to remember information from multiple sessions ago, making long-context robot memory important for real-world deployments. However, most robot-memory benchmarks today are based on single episodes or a short context. To measure how current robot memory systems perform on longer sessions wit…
12:00arXiv cs.ROOriginal ↗
VLA / Models Reinforcement Learning Research
ScaRF-SLAM: Scale-Consistent Reconstruction with Feed-Forward Models and Classical Visual SLAM
arXiv:2606.00307v2 Announce Type: replace Abstract: Recent works have explored unifying SLAM with geometric foundation models (GFMs). However, directly using GFM predictions for tracking is highly sensitive to model capability and uncertainty, as geometric inaccuracies in the predictions can adversely affect pose estimation. To address this limitation, we present a decoupled framework that integrates classical feature-based SLAM with GFMs, which achieves higher quality and more consistent dense …
12:00arXiv cs.ROOriginal ↗
Research
Symmetries Here and There, Combined Everywhere: Cross-space Symmetry Compositions in Robotics
arXiv:2605.22639v3 Announce Type: replace Abstract: Robots exhibit a rich variety of symmetries arising from their mechanical structure and the properties of their tasks. Although many robotics problems exhibit several symmetries simultaneously, existing approaches typically treat them in isolation, failing to exploit their combined potential. This paper introduces cross-space symmetry compositions, a framework for learning robot policies that are jointly equivariant to multiple symmetries acros…
12:00arXiv cs.ROOriginal ↗
Research
Minimalist Visual Inertial Odometry
arXiv:2605.19990v2 Announce Type: replace Abstract: Visual-Inertial Odometry (VIO), which is critical to mobile robot navigation, uses cameras with a large number of pixels. Capturing and processing camera images requires significant resources. This work presents a minimalist approach to planar odometry, showing that just four visual sensors (pixels) and an IMU provide robust motion estimation for differential-drive robots. Our key insight is that four downward-facing photodiodes that sense the …
12:00arXiv cs.ROOriginal ↗
Research
Tutorial on Aided Inertial Navigation Systems: A Modern Treatment Using Lie-Group Theoretical Methods
arXiv:2603.07143v2 Announce Type: replace Abstract: This tutorial presents a control-oriented introduction to aided inertial navigation systems using a Lie-group formulation centered on the extended Special Euclidean group SE_2(3). The focus is on developing a clear and implementation-oriented geometric framework for fusing inertial measurements with aiding information, while making the role of invariance and symmetry explicit. Recent extensions, including higher-order state representations, syn…
12:00arXiv cs.ROOriginal ↗
Research
MI-SLAM: Magnetic Inertial SLAM Systems
arXiv:2512.10128v4 Announce Type: replace Abstract: Spatially inhomogeneous magnetic fields offer a valuable, non-visual information source for positioning. Among systems leveraging this, magnetic field-based simultaneous localization and mapping (SLAM) systems are particularly attractive. These systems execute positioning and magnetic field mapping tasks simultaneously, and they have bounded positioning error within previously visited regions. However, state-of-the-art magnetic-field SLAM metho…
12:00arXiv cs.ROOriginal ↗
Reinforcement Learning Research
Performance-guided Task-specific Optimization for Multirotor Design
arXiv:2510.04724v2 Announce Type: replace Abstract: This paper introduces a methodology for task-specific design optimization of multirotor Micro Aerial Vehicles. By leveraging reinforcement learning, Bayesian optimization, and covariance matrix adaptation evolution strategy, we optimize aerial robot designs guided exclusively by their closed-loop performance in a considered task. Our approach systematically explores the design space of motor pose configurations while ensuring manufacturability …
12:00arXiv cs.ROOriginal ↗

Industry 3185 articles · 574 matching

Aug 28 Fri5 articles

Aug 27 Thu25 articles