{"@context":"https://schema.org","@type":"NewsArticle","generatedAt":"2026-07-23T06:40:50.084Z","headline":"小米发布机器人策略模型 Xiaomi-Robotics-1","description":"小米推出 Xiaomi-Robotics-1，结合大规模无具身（UMI）预训练与少量真实机器人数据后训练。预训练使用 10 万小时、覆盖 1700+ 场景的 UMI 轨迹数据，后训练引入 7200+ 小时真实家庭数据。模型在四项仿真基准上取得 SOTA，新任务平均不到 10 小时演示即可达 75% 成功率。","url":"https://www.aioga.com/news/cmrsydl3b056vbitl77uxf5rp/","mainEntityOfPage":"https://www.aioga.com/news/cmrsydl3b056vbitl77uxf5rp/","datePublished":"2026-07-20T07:52:12.469Z","dateModified":"2026-07-20T07:52:12.469Z","inLanguage":"zh-CN","publisher":{"@type":"NewsMediaOrganization","name":"Aioga","url":"https://www.aioga.com"},"citation":["https://robotics.xiaomi.com/xiaomi-robotics-1.html","https://aihot.virxact.com/items/cmrsydl3b056vbitl77uxf5rp"],"canonicalUrl":"https://www.aioga.com/news/cmrsydl3b056vbitl77uxf5rp/","directAnswer":{"@type":"Answer","text":"Aioga 编辑摘要：小米推出 Xiaomi-Robotics-1，结合大规模无具身（UMI）预训练与少量真实机器人数据后训练。 Aioga 将其归入「模型更新」方向，重点关注它对真实使用和行业竞争的影响。","url":"https://www.aioga.com/news/cmrsydl3b056vbitl77uxf5rp/","dateCreated":"2026-07-20T07:52:12.469Z","author":{"@type":"Organization","@id":"https://www.aioga.com/authors/aioga-editorial/#editorial-team","name":"Aioga Editorial Team","url":"https://www.aioga.com/authors/aioga-editorial/"}},"evidence":[{"@type":"CreativeWork","name":"robotics.xiaomi.com source article","url":"https://robotics.xiaomi.com/xiaomi-robotics-1.html","datePublished":"2026-07-20T07:52:12.469Z","provider":{"@type":"Organization","name":"robotics.xiaomi.com","url":"https://robotics.xiaomi.com/xiaomi-robotics-1.html"}},{"@type":"CreativeWork","name":"AIHot archive record","url":"https://aihot.virxact.com/items/cmrsydl3b056vbitl77uxf5rp","datePublished":"2026-07-20T07:52:12.469Z","provider":{"@type":"Organization","name":"AIHot","url":"https://aihot.virxact.com/items/cmrsydl3b056vbitl77uxf5rp"}}],"aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","originalPublisher":{"name":"robotics.xiaomi.com","url":"https://robotics.xiaomi.com/xiaomi-robotics-1.html"},"article":{"id":"cmrsydl3b056vbitl77uxf5rp","slug":"cmrsydl3b056vbitl77uxf5rp","url":"https://www.aioga.com/news/cmrsydl3b056vbitl77uxf5rp/","title":"小米发布机器人策略模型 Xiaomi-Robotics-1","title_en":"小米-机器人-1","summary":"小米推出 Xiaomi-Robotics-1，结合大规模无具身（UMI）预训练与少量真实机器人数据后训练。预训练使用 10 万小时、覆盖 1700+ 场景的 UMI 轨迹数据，后训练引入 7200+ 小时真实家庭数据。模型在四项仿真基准上取得 SOTA，新任务平均不到 10 小时演示即可达 75% 成功率。","source":"Hacker News 热门（buzzing.cc 中文翻译）","sourceUrl":"https://robotics.xiaomi.com/xiaomi-robotics-1.html","aiHotUrl":"https://aihot.virxact.com/items/cmrsydl3b056vbitl77uxf5rp","publishedAt":"2026-07-20T07:52:12.469Z","category":"模型更新","score":66,"selected":false,"articleBody":["Xiaomi-Robotics-1 is a ready-to-use robot foundation model trained on over 100K hours of real-world manipulation trajectories.","Breaking the data barrier. Scaling robot policy models with embodiment-free pre-training.","Foundation models in language and vision keep moving the frontier by riding empirical scaling laws: capability tracks data, parameters, and compute. Robotics has missed out. Large-scale, high-quality data is hard to come by, and that scarcity, more than anything else, has capped how far policy models could scale. What robots can do under genuinely large-scale training remained largely an open question. We take a step toward answering it. Xiaomi-Robotics-1 combines large-scale embodiment-free (UMI) pre-training with a modest amount of real-robot data in a post-training stage. We study how the model behaves as it scales.","Everything Xiaomi-Robotics-1 can do starts from data. For pre-training, we use 100,000 hours of embodiment-free (UMI) trajectories spanning more than 1,700 scenarios (household, commercial premises, industrial sites, and outdoor spaces), covering a diverse range of tasks. We develop a scalable auto-labeling pipeline that first divides trajectories into fixed-length segments and then annotates each segment with language descriptions of scene state transitions.","For post-training, we leverage cross-embodiment datasets containing in-house robot data, filtered open-sourced robot data, and a set of high-quality UMI data. For the in-house data, we collected over 7,200 hours of real-robot data in real homes, covering tasks like tidying a sofa, sorting a shoe cabinet, and putting away kitchenware. The UMI data are manually annotated with temporal segments and instruction prompts, which differ from the auto-labeled state-transition descriptions used in the pre-training data.","Following the training paradigm of LLMs, the training of Xiaomi-Robotics-1 consists of two stages: pre-training and post-training. The first stage learns general representations for action generation from large-scale UMI data, while the post-training stage aligns the model with real robot embodiments and instruction-following capabilities.","Pre-training is about breadth: exposing the model to as much of the real world as possible. We use the embodiment-free UMI data described above, which spans a broad range of environments and tasks. At this scale, manual labeling is infeasible. Thus, we built an automatic annotation pipeline powered by a strong vision-language model. Long videos are split into fixed-length clips, and the VLM describes the state transition of grippers and interacting objects within each clip. The result is a large-scale corpus of real-world manipulation trajectories, each annotated with precise language descriptions. These allow the model to learn action generation that drives the scene toward the state transitions described by the language.","An encouraging finding is that pre-training shows a clean scaling behavior: as data and model size grow, validation action error steadily decreases.","Post-training aims to align the strong action-generation capabilities acquired from pre-training with real robot embodiments and natural-language instruction following along two axes. Embodiment alignment uses high-quality cross-embodiment real-robot data to map the general action-generation ability onto actual robots. Instruction alignment shifts the model from \"generating actions given a description of scene state transitions\" to \"understanding a natural-language instruction and executing it directly.\"","After post-training, Xiaomi-Robotics-1 can be used out-of-the-box to perform a wide range of mobile manipulation tasks in the real world. We evaluate the post-trained model in unseen environments with unseen object instances to understand whether the scaling behaviors from pre-training can transfer to real-robot performance after post-training.","The answer is yes. As we increase the amount of pre-training data and model size, real-robot success rate rises steadily and predictably. That is, a stronger pre-trained model yields better real-robot performance. The scaling gains show no signs of saturation: the real-robot success rate after post-training keeps improving as the model consumes more data or scales up during pre-training.","After post-training, Xiaomi-Robotics-1 can serve as a strong robot foundation model for downstream applications. We put Xiaomi-Robotics-1 to use in two complementary downstream settings. Efficient adaptation to new tasks specializes the model to brand-new, highly complex real-robot tasks from a few hours of data per task. Simulation benchmarks probe its capabilities in mainstream suites that emphasize generalization.","Xiaomi-Robotics-1 can learn new tasks with high data efficiency. The model picks up tasks like phone packing, printer refilling, laundry loading, and box packing from just a few hours of real-robot demonstrations per task. With an average of under 10 hours of demonstrations per task, it already reaches a 75% overall success rate, nearly doubling the π 0.5 baseline (40%) at the same budget; raising the budget to an average of under 40 hours lifts overall success to 85%.","Evaluation on efficient learning of new tasks. Each cell shows success rate (%), higher is better. XR-1 = Xiaomi-Robotics-1.","We evaluate Xiaomi-Robotics-1 on four mainstream simulation benchmarks. It achieves state-of-the-art results on all four benchmarks. The table reports the average success rate and the relative gain over second place. These results show that the generalization and scaling gains of Xiaomi-Robotics-1 carry over to standard simulation evaluation.","Simulation evaluation. All benchmarks report average success rate (%). XR-1 = Xiaomi-Robotics-1; Rel. Gain = (XR-1 − 2nd best) / 2nd best. Higher is better.","Xiaomi-Robotics-1 demonstrates a practical path for scaling robot foundation models: large-scale embodiment-free UMI pre-training breaks the robot data bottleneck, while real-robot and instruction alignment transfer that general capability to physical robots. Results show that the model scales neatly with data volume and model size during pre-training, and that this scaling behavior translates directly to post-training, where a stronger pre-trained model yields better out-of-the-box real-robot performance in unseen environments. The resulting foundation model adapts to new tasks from minimal data and achieves state-of-the-art performance on four challenging simulation benchmarks that emphasize generalization.","Finally, we present an uncut footage of luggage packing."],"articleImages":[{"sourceUrl":"https://robotics.xiaomi.com/robot-static-resource/xiaomi-robotics-1/pretrain_data.jpg","alt":"Pretrain data overview","afterParagraph":3,"url":"/media/articles/cmrsydl3b056vbitl77uxf5rp/dd685b5e7ed4dcf1.jpg"},{"sourceUrl":"https://robotics.xiaomi.com/robot-static-resource/xiaomi-robotics-1/posttrain_data.jpg","alt":"Posttrain data overview","afterParagraph":4,"url":"/media/articles/cmrsydl3b056vbitl77uxf5rp/8357b7272923deab.jpg"},{"sourceUrl":"https://robotics.xiaomi.com/robot-static-resource/xiaomi-robotics-1/pretrain_scaling.jpg","alt":"Pretrain scaling curve","afterParagraph":7,"url":"/media/articles/cmrsydl3b056vbitl77uxf5rp/d62da752fb29bf2a.jpg"},{"sourceUrl":"https://robotics.xiaomi.com/robot-static-resource/xiaomi-robotics-1/posttrain_scaling.jpg","alt":"Post-training scaling: real-robot success rate vs data ratio and model size","afterParagraph":10,"url":"/media/articles/cmrsydl3b056vbitl77uxf5rp/197d731792c3638b.jpg"},{"sourceUrl":"https://robotics.xiaomi.com/robot-static-resource/xiaomi-robotics-1/robocasa365.jpg","alt":"RoboCasa365 leaderboard","afterParagraph":15,"url":"/media/articles/cmrsydl3b056vbitl77uxf5rp/92640b4952ec6666.jpg"},{"sourceUrl":"https://robotics.xiaomi.com/robot-static-resource/xiaomi-robotics-1/robodojo.jpg","alt":"RoboDojo leaderboard","afterParagraph":15,"url":"/media/articles/cmrsydl3b056vbitl77uxf5rp/e3c249cda2405ee1.jpg"}],"mediaStatus":"ok","articleBodyZh":["小米机器人-1 是一个可直接使用的机器人基础模型，经过超过 10 万小时的真实操作轨迹训练。","打破数据障碍。通过无实体预训练扩展机器人策略模型。","语言和视觉领域的基础模型通过遵循经验规模规律不断推动前沿发展：能力随数据、参数和算力增长。机器人领域却落后了。大规模、高质量的数据难以获得，而这种稀缺，比其他任何因素都更限制了策略模型的扩展能力。机器人在真正大规模训练下能做到什么仍然基本是个未解之谜。我们迈出了一步来回答这一问题。小米机器人-1 结合了大规模无实体（UMI）预训练和少量真实机器人数据的后训练阶段。我们研究了模型随规模增长的表现。","小米机器人-1 能做的所有事情都始于数据。在预训练中，我们使用了 10 万小时的无实体（UMI）轨迹，涵盖超过 1,700 种场景（家庭、商业场所、工业现场和户外空间），涉及各种任务。我们开发了一个可扩展的自动标注流程，首先将轨迹划分为固定长度的片段，然后为每个片段注释场景状态变化的语言描述。","在后训练阶段，我们利用了跨实体数据集，包括自有机器人数据、筛选的开源机器人数据以及一套高质量的 UMI 数据。对于自有数据，我们在真实家庭环境中收集了超过 7,200 小时的真实机器人数据，涵盖整理沙发、整理鞋柜、收纳厨具等任务。UMI 数据经过人工标注，包含时间片段和操作指令提示，不同于预训练数据中使用的自动标注状态变化描述。","遵循大型语言模型的训练范式，小米机器人-1 的训练包括两个阶段：预训练和后训练。第一阶段从大规模 UMI 数据中学习行动生成的一般表示，后训练阶段则将模型与真实机器人实体和指令执行能力对齐。","预训练关注的是广度：让模型尽可能多地接触现实世界。我们使用上述无体态 UMI 数据，该数据涵盖广泛的环境和任务。在这个规模下，人工标注是不可行的。因此，我们构建了一个由强大的视觉-语言模型驱动的自动标注流程。长视频被分割成固定长度的片段，VLM 描述每个片段中抓手和交互物体的状态变化。结果是大规模的现实世界操作轨迹语料库，每条轨迹都配有精确的语言描述。这使得模型能够学习动作生成，从而推动场景向语言描述的状态转换发展。","一个令人鼓舞的发现是，预训练显示出清晰的规模效应：随着数据量和模型规模的增长，验证动作误差稳步下降。","后训练的目标是沿两个轴线将预训练中获得的强大动作生成能力与真实机器人体态及自然语言指令执行对齐。体态对齐使用高质量的跨体态真实机器人数据，将通用的动作生成能力映射到实际机器人上。指令对齐则将模型从“根据场景状态变化的描述生成动作”转变为“理解自然语言指令并直接执行”。","经过后训练，Xiaomi-Robotics-1 可以开箱即用地在现实世界中执行各种移动操作任务。我们在未见过的环境和未见过的物体实例中评估后训练模型，以了解预训练的规模效应是否可以在后训练后转移到真实机器人性能上。","答案是肯定的。随着预训练数据量和模型规模的增加，真实机器人成功率稳步且可预测地上升。也就是说，更强的预训练模型带来更好的真实机器人性能。规模效应的收益没有显示出饱和迹象：后训练后的真实机器人成功率随着模型在预训练期间消耗更多数据或扩大规模而持续改善。","经过后续训练，Xiaomi-Robotics-1 可以作为下游应用的强大机器人基础模型。我们在两个互补的下游场景中应用了 Xiaomi-Robotics-1。对新任务的高效适应将模型专门化到全新的、高度复杂的真实机器人任务，每个任务仅需几小时的数据。模拟基准测试则在强调泛化能力的主流套件中检验其能力。","Xiaomi-Robotics-1 可以以高数据效率学习新任务。该模型仅通过每个任务几小时的真实机器人演示，就能掌握如手机打包、打印机加墨、洗衣机装载和箱子打包等任务。在每个任务平均不到 10 小时的演示下，它的整体成功率已达到 75%，几乎是相同预算下 π 0.5 基线（40%）的两倍；将预算提高到平均不到 40 小时，总体成功率可提升到 85%。","新任务高效学习的评估。每个单元格显示成功率（%），数值越高越好。XR-1 = Xiaomi-Robotics-1。","我们在四个主流模拟基准上评估 Xiaomi-Robotics-1。它在所有四个基准中都取得了最先进的结果。表格报告了平均成功率及相对于第二名的提升比例。这些结果表明，Xiaomi-Robotics-1 的泛化能力和规模收益可延展到标准模拟评估。","模拟评估。所有基准均报告平均成功率（%）。XR-1 = Xiaomi-Robotics-1；相对增益 = (XR-1 − 第二名) / 第二名。数值越高越好。","Xiaomi-Robotics-1 展示了扩展机器人基础模型的实用路径：大规模、无实体约束的 UMI 预训练打破了机器人数据瓶颈，而真实机器人和指令对齐将这种泛化能力转移到物理机器人。结果显示，模型在预训练期间随数据量和模型规模呈现良好扩展性，并且这种扩展行为可以直接转化到后续训练中，在未见过的环境中，预训练更强的模型能够提供更好的开箱即用真实机器人性能。最终的基础模型能够从最少的数据适应新任务，并在四个强调泛化能力的挑战性模拟基准上达到最先进性能。","最后，我们展示了行李打包的未剪辑视频。"],"translationStatus":"translated","bodyOrigin":"source-page","editorial":{"summary":"Aioga 编辑摘要：小米推出 Xiaomi-Robotics-1，结合大规模无具身（UMI）预训练与少量真实机器人数据后训练。 Aioga 将其归入「模型更新」方向，重点关注它对真实使用和行业竞争的影响。","background":"背景分析：模型与研究类动态需要结合能力边界、开放方式、成本、可用性和真实任务表现判断，单项指标领先不等于已经形成稳定采用。","viewpoint":"Aioga 判断：这条动态更适合作为行业观察信号，当前信息足以建立线索，但不足以推导长期结论。","implications":"影响分析：对相关团队而言，短期应先核对来源、可用范围和实际成本，再判断是否值得接入或跟进。","nextStep":"后续观察：继续观察官方文档、实际可用性、价格变化、开发者反馈和竞品回应。","evidenceRefs":["title","summary","articleBody"],"confidence":"medium","status":"published","aiGenerated":false,"autoApproved":true,"generatedBy":"rule-safe-fallback","generatedAt":"2026-07-23T06:49:19.114Z","sourceHash":"1a6d89b8c120a583","validation":{"passed":true,"mode":"rule-safe-fallback","checks":["schema","length","source-attribution","no-html"]}},"tags":["模型更新","Hacker News 热门（buzzing.cc 中文翻译）"],"translations":{"zh-CN":{"title":"小米发布机器人策略模型 Xiaomi-Robotics-1","summary":"小米推出 Xiaomi-Robotics-1，结合大规模无具身（UMI）预训练与少量真实机器人数据后训练。预训练使用 10 万小时、覆盖 1700+ 场景的 UMI 轨迹数据，后训练引入 7200+ 小时真实家庭数据。模型在四项仿真基准上取得 SOTA，新任务平均不到 10 小时演示即可达 75% 成功率。","category":"模型更新","source":"robotics.xiaomi.com","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"小米发布机器人策略模型 Xiaomi-Robotics-1 - Aioga AI资讯","description":"小米推出 Xiaomi-Robotics-1，结合大规模无具身（UMI）预训练与少量真实机器人数据后训练。预训练使用 10 万小时、覆盖 1700+ 场景的 UMI 轨迹数据，后训练引入 7200+ 小时真实家庭数据。模型在四项仿真基准上取得 SOTA，新任务平均不到 10 小时演示即可达 75% 成功率。","url":"https://www.aioga.com/news/cmrsydl3b056vbitl77uxf5rp/"},"en":{"title":"Xiaomi releases robot strategy model Xiaomi-Robotics-1","summary":"Xiaomi launched Xiaomi-Robotics-1, which is trained by combining large-scale unembodied (UMI) pretraining with a small amount of real robot data. The pretraining uses 100,000 hours of UMI trajectory data covering more than 1,700 scenarios, and the subsequent training introduces over 7,200 hours of real household data. The model achieves state-of-the-art (SOTA) results on four simulation benchmarks, and for new tasks, it can reach a 75% success rate with an average of less than 10 hours of demonstrations.","category":"Models","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"Xiaomi releases robot strategy model Xiaomi-Robotics-1 - Aioga AI News","description":"Xiaomi launched Xiaomi-Robotics-1, which is trained by combining large-scale unembodied (UMI) pretraining with a small amount of real robot data. The pretraining uses 100,000 hours...","url":"https://www.aioga.com/en/news/cmrsydl3b056vbitl77uxf5rp/","contentTranslated":true,"sourceHash":"88f9ed5cf2146ffd","translatedAt":"2026-07-22T19:24:06.791Z"},"ja":{"title":"小米、ロボット戦略モデル Xiaomi-Robotics-1 を発表","summary":"Xiaomi は Xiaomi-Robotics-1 を発表し、大規模な非具現（UMI）事前学習と少量の実際のロボットデータを組み合わせて学習しました。事前学習には 10 万時間、1700 以上のシーンをカバーする UMI 軌跡データを使用し、その後の学習では 7200 時間以上の実際の家庭データを導入しました。モデルは 4 つのシミュレーションベンチマークで SOTA を達成し、新しいタスクでは平均で 10 時間未満のデモで 75% の成功率に到達しました。","category":"モデル更新","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"小米、ロボット戦略モデル Xiaomi-Robotics-1 を発表 - Aioga AIニュース","description":"Xiaomi は Xiaomi-Robotics-1 を発表し、大規模な非具現（UMI）事前学習と少量の実際のロボットデータを組み合わせて学習しました。事前学習には 10 万時間、1700 以上のシーンをカバーする UMI 軌跡データを使用し、その後の学習では 7200 時間以上の実際の家庭データを導入しました。モデルは 4 つのシミュレーションベンチマーク...","url":"https://www.aioga.com/ja/news/cmrsydl3b056vbitl77uxf5rp/","contentTranslated":true,"sourceHash":"88f9ed5cf2146ffd","translatedAt":"2026-07-22T19:24:05.865Z"},"ko":{"title":"샤오미, 로봇 전략 모델 Xiaomi-Robotics-1 출시","summary":"샤오미는 Xiaomi-Robotics-1을 출시했으며, 대규모 비체험(UMI) 사전학습과 소량의 실제 로봇 데이터를 결합한 후 학습한다. 사전학습에는 10만 시간, 1700개 이상의 장면을 포함한 UMI 궤적 데이터가 사용되며, 이후 학습에는 7200시간 이상의 실제 가정 데이터가 도입된다. 모델은 네 가지 시뮬레이션 벤치마크에서 SOTA를 달성했으며, 새로운 과제에서는 평균 10시간도 채 되지 않는 시연으로 75% 성공률에 도달한다.","category":"모델 업데이트","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"샤오미, 로봇 전략 모델 Xiaomi-Robotics-1 출시 - Aioga AI 뉴스","description":"샤오미는 Xiaomi-Robotics-1을 출시했으며, 대규모 비체험(UMI) 사전학습과 소량의 실제 로봇 데이터를 결합한 후 학습한다. 사전학습에는 10만 시간, 1700개 이상의 장면을 포함한 UMI 궤적 데이터가 사용되며, 이후 학습에는 7200시간 이상의 실제 가정 데이터가 도입된다. 모델은 네 가지 시뮬레이션 벤...","url":"https://www.aioga.com/ko/news/cmrsydl3b056vbitl77uxf5rp/","contentTranslated":true,"sourceHash":"88f9ed5cf2146ffd","translatedAt":"2026-07-22T19:24:58.826Z"},"es":{"title":"Xiaomi lanza el modelo de estrategia de robots Xiaomi-Robotics-1","summary":"Xiaomi lanzó Xiaomi-Robotics-1, entrenado combinando un preentrenamiento a gran escala sin encarnación (UMI) con una pequeña cantidad de datos de robots reales. El preentrenamiento utilizó 100,000 horas de datos de trayectorias UMI que cubren más de 1,700 escenarios, y el entrenamiento posterior incorporó más de 7,200 horas de datos reales del hogar. El modelo logró el SOTA en cuatro puntos de referencia de simulación, y para nuevas tareas, se necesita en promedio menos de 10 horas de demostración para alcanzar una tasa de éxito del 75%.","category":"Modelos","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"Xiaomi lanza el modelo de estrategia de robots Xiaomi-Robotics-1 - Aioga Noticias de IA","description":"Xiaomi lanzó Xiaomi-Robotics-1, entrenado combinando un preentrenamiento a gran escala sin encarnación (UMI) con una pequeña cantidad de datos de robots reales. El preentrenamiento...","url":"https://www.aioga.com/es/news/cmrsydl3b056vbitl77uxf5rp/","contentTranslated":true,"sourceHash":"88f9ed5cf2146ffd","translatedAt":"2026-07-22T19:24:48.748Z"},"fr":{"title":"Xiaomi publie le modèle stratégique de robotique Xiaomi-Robotics-1","summary":"Xiaomi a lancé Xiaomi-Robotics-1, entraîné en combinant une pré-formation à grande échelle sans corps (UMI) avec une petite quantité de données provenant de vrais robots. La pré-formation a utilisé 100 000 heures de données de trajectoire UMI couvrant plus de 1 700 scènes, et l'entraînement ultérieur a introduit plus de 7 200 heures de données réelles issues de foyers. Le modèle a atteint l'état de l'art sur quatre benchmarks de simulation, et pour les nouvelles tâches, moins de 10 heures de démonstration suffisent en moyenne pour atteindre un taux de réussite de 75 %.","category":"Modèles","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"Xiaomi publie le modèle stratégique de robotique Xiaomi-Robotics-1 - Aioga Actualités IA","description":"Xiaomi a lancé Xiaomi-Robotics-1, entraîné en combinant une pré-formation à grande échelle sans corps (UMI) avec une petite quantité de données provenant de vrais robots. La pré-fo...","url":"https://www.aioga.com/fr/news/cmrsydl3b056vbitl77uxf5rp/","contentTranslated":true,"sourceHash":"88f9ed5cf2146ffd","translatedAt":"2026-07-22T19:25:39.914Z"},"de":{"title":"Xiaomi stellt Robotik-Strategiemodell Xiaomi-Robotics-1 vor","summary":"Xiaomi hat den Xiaomi-Robotics-1 vorgestellt, der nach einer groß angelegten unverkörperten (UMI) Vortrainierung und mit einer kleinen Menge an echten Roboterdaten trainiert wurde. Das Vortraining verwendete 100.000 Stunden UMI-Trajektoriendaten, die über 1.700 Szenen abdecken, und das Nachtraining führte mehr als 7.200 Stunden echte Haushaltsdaten ein. Das Modell erzielte SOTA-Ergebnisse in vier Simulationsbenchmarks, und bei neuen Aufgaben reicht im Durchschnitt weniger als 10 Stunden Demonstration, um eine Erfolgsrate von 75 % zu erreichen.","category":"模型更新","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"Xiaomi stellt Robotik-Strategiemodell Xiaomi-Robotics-1 vor - Aioga KI-News","description":"Xiaomi hat den Xiaomi-Robotics-1 vorgestellt, der nach einer groß angelegten unverkörperten (UMI) Vortrainierung und mit einer kleinen Menge an echten Roboterdaten trainiert wurde....","url":"https://www.aioga.com/de/news/cmrsydl3b056vbitl77uxf5rp/","contentTranslated":true,"sourceHash":"88f9ed5cf2146ffd","translatedAt":"2026-07-22T19:25:48.669Z"},"pt-BR":{"title":"Xiaomi lança modelo de estratégia de robótica Xiaomi-Robotics-1","summary":"A Xiaomi lançou o Xiaomi-Robotics-1, treinado combinando pré-treinamento em Larga Escala sem Corpo (UMI) com uma pequena quantidade de dados reais de robôs. O pré-treinamento utilizou 100 mil horas de dados de trajetórias UMI cobrindo mais de 1700 cenários, e o pós-treinamento incorporou mais de 7200 horas de dados reais de residências. O modelo alcançou SOTA em quatro benchmarks de simulação, e em tarefas novas requer, em média, menos de 10 horas de demonstrações para atingir uma taxa de sucesso de 75%.","category":"模型更新","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"Xiaomi lança modelo de estratégia de robótica Xiaomi-Robotics-1 - Aioga Notícias de IA","description":"A Xiaomi lançou o Xiaomi-Robotics-1, treinado combinando pré-treinamento em Larga Escala sem Corpo (UMI) com uma pequena quantidade de dados reais de robôs. O pré-treinamento utili...","url":"https://www.aioga.com/pt-BR/news/cmrsydl3b056vbitl77uxf5rp/","contentTranslated":true,"sourceHash":"88f9ed5cf2146ffd","translatedAt":"2026-07-22T19:26:31.943Z"},"ru":{"title":"Xiaomi выпускает стратегическую модель робота Xiaomi-Robotics-1","summary":"Xiaomi выпустила Xiaomi-Robotics-1, которая обучается с использованием масштабного предварительного обучения без физического тела (UMI) и небольшого объема данных настоящих роботов. Предварительное обучение использует 100 000 часов данных траекторий UMI, охватывающих более 1700 сценариев, после чего обучение продолжается с использованием более 7200 часов данных из реальных домашних условий. Модель достигла SOTA на четырех симуляционных эталонах, а для новых задач в среднем требуется менее 10 часов демонстраций, чтобы достичь 75% успешности.","category":"模型更新","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"Xiaomi выпускает стратегическую модель робота Xiaomi-Robotics-1 - Aioga Новости ИИ","description":"Xiaomi выпустила Xiaomi-Robotics-1, которая обучается с использованием масштабного предварительного обучения без физического тела (UMI) и небольшого объема данных настоящих роботов...","url":"https://www.aioga.com/ru/news/cmrsydl3b056vbitl77uxf5rp/","contentTranslated":true,"sourceHash":"88f9ed5cf2146ffd","translatedAt":"2026-07-22T19:26:30.578Z"},"ar":{"title":"شاومي تطلق نموذج استراتيجية الروبوتات Xiaomi-Robotics-1","summary":"أطلقت شياومي جهاز Xiaomi-Robotics-1، والذي يجمع بين التدريب المسبق واسع النطاق بدون جسم (UMI) وبعدها التدريب باستخدام كمية صغيرة من بيانات الروبوت الحقيقية. استخدم التدريب المسبق 100 ألف ساعة من بيانات مسار UMI التي تغطي أكثر من 1700 مشهد، بينما أُدخل التدريب اللاحق 7200 ساعة من بيانات المنازل الحقيقية. حقق النموذج نتائج SOTA في أربعة معايير محاكاة، ويمكنه في المهام الجديدة الوصول لمعدل نجاح 75٪ بعد أقل من 10 ساعات من العروض التجريبية في المتوسط.","category":"模型更新","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"شاومي تطلق نموذج استراتيجية الروبوتات Xiaomi-Robotics-1 - Aioga أخبار الذكاء الاصطناعي","description":"أطلقت شياومي جهاز Xiaomi-Robotics-1، والذي يجمع بين التدريب المسبق واسع النطاق بدون جسم (UMI) وبعدها التدريب باستخدام كمية صغيرة من بيانات الروبوت الحقيقية. استخدم التدريب المسبق 1...","url":"https://www.aioga.com/ar/news/cmrsydl3b056vbitl77uxf5rp/","contentTranslated":true,"sourceHash":"88f9ed5cf2146ffd","translatedAt":"2026-07-22T19:27:21.195Z"},"hi":{"title":"शाओमी ने रोबोटिक्स रणनीति मॉडल Xiaomi-Robotics-1 जारी किया","summary":"Xiaomi ने Xiaomi-Robotics-1 लॉन्च किया, जो बड़े पैमाने पर अव्यक्त (UMI) पूर्व-प्रशिक्षण और थोड़े वास्तविक रोबोट डेटा के संयोजन के बाद प्रशिक्षित किया गया। पूर्व-प्रशिक्षण में 10 लाख घंटे, 1700+ दृश्य को कवर करने वाले UMI ट्राजेक्टरी डेटा का उपयोग किया गया, और बाद के प्रशिक्षण में 7200+ घंटे का वास्तविक घरेलू डेटा शामिल किया गया। मॉडल ने चार सिमुलेशन बेन्चमार्क पर SOTA हासिल किया, और नए कार्यों में औसतन केवल 10 घंटे के प्रदर्शन में 75% सफलता दर प्राप्त की।","category":"模型更新","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"शाओमी ने रोबोटिक्स रणनीति मॉडल Xiaomi-Robotics-1 जारी किया - Aioga AI समाचार","description":"Xiaomi ने Xiaomi-Robotics-1 लॉन्च किया, जो बड़े पैमाने पर अव्यक्त (UMI) पूर्व-प्रशिक्षण और थोड़े वास्तविक रोबोट डेटा के संयोजन के बाद प्रशिक्षित किया गया। पूर्व-प्रशिक्षण में 10 ला...","url":"https://www.aioga.com/hi/news/cmrsydl3b056vbitl77uxf5rp/","contentTranslated":true,"sourceHash":"88f9ed5cf2146ffd","translatedAt":"2026-07-22T19:27:21.840Z"},"it":{"title":"Xiaomi ha rilasciato il modello strategico dei robot Xiaomi-Robotics-1","summary":"Xiaomi ha lanciato Xiaomi-Robotics-1, addestrato combinando un pre-addestramento su larga scala senza corpo (UMI) con una piccola quantità di dati provenienti da robot reali. Il pre-addestramento ha utilizzato 100.000 ore di dati di traiettoria UMI coprendo oltre 1.700 scenari, mentre l'addestramento successivo ha introdotto oltre 7.200 ore di dati reali provenienti da abitazioni. Il modello ha raggiunto lo stato dell'arte (SOTA) su quattro benchmark di simulazione e per nuovi compiti sono sufficienti in media meno di 10 ore di dimostrazioni per raggiungere un tasso di successo del 75%.","category":"模型更新","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"Xiaomi ha rilasciato il modello strategico dei robot Xiaomi-Robotics-1 - Aioga Notizie IA","description":"Xiaomi ha lanciato Xiaomi-Robotics-1, addestrato combinando un pre-addestramento su larga scala senza corpo (UMI) con una piccola quantità di dati provenienti da robot reali. Il pr...","url":"https://www.aioga.com/it/news/cmrsydl3b056vbitl77uxf5rp/","contentTranslated":true,"sourceHash":"88f9ed5cf2146ffd","translatedAt":"2026-07-22T19:28:03.635Z"},"nl":{"title":"Xiaomi introduceert robotstrategiemodel Xiaomi-Robotics-1","summary":"Xiaomi heeft de Xiaomi-Robotics-1 gelanceerd, die wordt getraind door grootschalige onembodied (UMI) pre-training te combineren met een kleine hoeveelheid echte robotgegevens. De pre-training maakt gebruik van 100.000 uur aan UMI-trajectgegevens die meer dan 1.700 scenario's bestrijken, en in de latere training worden meer dan 7.200 uur aan echte huishoudgegevens toegevoegd. Het model behaalt SOTA-prestaties op vier simulatiebenchmarks en voor nieuwe taken is gemiddeld minder dan 10 uur demonstratie nodig om een succesratio van 75% te bereiken.","category":"模型更新","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"Xiaomi introduceert robotstrategiemodel Xiaomi-Robotics-1 - Aioga AI-nieuws","description":"Xiaomi heeft de Xiaomi-Robotics-1 gelanceerd, die wordt getraind door grootschalige onembodied (UMI) pre-training te combineren met een kleine hoeveelheid echte robotgegevens. De p...","url":"https://www.aioga.com/nl/news/cmrsydl3b056vbitl77uxf5rp/","contentTranslated":true,"sourceHash":"88f9ed5cf2146ffd","translatedAt":"2026-07-22T19:28:02.714Z"},"tr":{"title":"Xiaomi, Robot Strateji Modeli Xiaomi-Robotics-1'i Tanıttı","summary":"Xiaomi, Xiaomi-Robotics-1'i tanıttı; bu sistem, büyük ölçekli bedenlenmemiş (UMI) ön eğitim ile az miktarda gerçek robot verisinin birleşimiyle eğitildi. Ön eğitim, 10 bin saatlik ve 1700'den fazla sahneyi kapsayan UMI yörünge verilerini kullanırken, ardından eğitime 7200'den fazla saatlik gerçek ev verisi dahil edildi. Model, dört simülasyon kıyaslamasında en iyi sonucu (SOTA) elde etti ve yeni görevlerde ortalama olarak 10 saatin altında bir gösterimle %75 başarı oranına ulaştı.","category":"模型更新","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"Xiaomi, Robot Strateji Modeli Xiaomi-Robotics-1'i Tanıttı - Aioga AI Haberleri","description":"Xiaomi, Xiaomi-Robotics-1'i tanıttı; bu sistem, büyük ölçekli bedenlenmemiş (UMI) ön eğitim ile az miktarda gerçek robot verisinin birleşimiyle eğitildi. Ön eğitim, 10 bin saatlik...","url":"https://www.aioga.com/tr/news/cmrsydl3b056vbitl77uxf5rp/","contentTranslated":true,"sourceHash":"88f9ed5cf2146ffd","translatedAt":"2026-07-22T19:28:52.097Z"},"vi":{"title":"Xiaomi phát hành mô hình chiến lược robot Xiaomi-Robotics-1","summary":"Xiaomi ra mắt Xiaomi-Robotics-1, kết hợp việc tiền huấn luyện không cơ thể quy mô lớn (UMI) với một lượng nhỏ dữ liệu robot thực tế sau khi huấn luyện. Tiền huấn luyện sử dụng 100.000 giờ dữ liệu quỹ đạo UMI, bao phủ hơn 1.700 cảnh. Sau khi huấn luyện, đưa vào 7.200+ giờ dữ liệu thực tế từ gia đình. Mô hình đạt SOTA trên bốn tiêu chuẩn mô phỏng, các nhiệm vụ mới trung bình chưa đến 10 giờ trình diễn đã đạt tỷ lệ thành công 75%.","category":"模型更新","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"Xiaomi phát hành mô hình chiến lược robot Xiaomi-Robotics-1 - Tin tức AI Aioga","description":"Xiaomi ra mắt Xiaomi-Robotics-1, kết hợp việc tiền huấn luyện không cơ thể quy mô lớn (UMI) với một lượng nhỏ dữ liệu robot thực tế sau khi huấn luyện. Tiền huấn luyện sử dụng 100....","url":"https://www.aioga.com/vi/news/cmrsydl3b056vbitl77uxf5rp/","contentTranslated":true,"sourceHash":"88f9ed5cf2146ffd","translatedAt":"2026-07-22T19:28:48.667Z"},"id":{"title":"Xiaomi merilis model strategi robot Xiaomi-Robotics-1","summary":"Xiaomi meluncurkan Xiaomi-Robotics-1, yang menggabungkan pretraining UMI (tanpa terwujud) berskala besar dengan sejumlah kecil data robot nyata untuk pelatihan selanjutnya. Pretraining menggunakan 100.000 jam data lintasan UMI yang mencakup lebih dari 1.700 skenario, kemudian pelatihan selanjutnya melibatkan lebih dari 7.200 jam data rumah nyata. Model ini mencapai SOTA pada empat tolok ukur simulasi, dan untuk tugas baru rata-rata hanya membutuhkan kurang dari 10 jam demonstrasi untuk mencapai tingkat keberhasilan 75%.","category":"模型更新","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"Xiaomi merilis model strategi robot Xiaomi-Robotics-1 - Berita AI Aioga","description":"Xiaomi meluncurkan Xiaomi-Robotics-1, yang menggabungkan pretraining UMI (tanpa terwujud) berskala besar dengan sejumlah kecil data robot nyata untuk pelatihan selanjutnya. Pretrai...","url":"https://www.aioga.com/id/news/cmrsydl3b056vbitl77uxf5rp/","contentTranslated":true,"sourceHash":"88f9ed5cf2146ffd","translatedAt":"2026-07-22T19:29:36.522Z"},"th":{"title":"เสียวหมี่เปิดตัวโมเดลกลยุทธ์หุ่นยนต์ Xiaomi-Robotics-1","summary":"Xiaomi เปิดตัว Xiaomi-Robotics-1 ซึ่งรวมการฝึกล่วงหน้าแบบไม่ต้องมีร่างกายจำนวนมาก (UMI) กับข้อมูลหุ่นยนต์จริงเพียงเล็กน้อย หลังจากการฝึกล่วงหน้า ใช้ข้อมูลร่องรอย UMI 100,000 ชั่วโมง ครอบคลุมมากกว่า 1,700 สถานการณ์ ในการฝึกขั้นตอนถัดมา มีการนำเข้าข้อมูลจริงในบ้านมากกว่า 7,200 ชั่วโมง โมเดลได้สร้างผลลัพธ์ SOTA ในสี่มาตรฐานการจำลองงาน และสำหรับงานใหม่ ๆ เฉลี่ยใช้เวลาน้อยกว่า 10 ชั่วโมงในการสาธิตก็สามารถบรรลุอัตราความสำเร็จ 75%","category":"模型更新","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"เสียวหมี่เปิดตัวโมเดลกลยุทธ์หุ่นยนต์ Xiaomi-Robotics-1 - ข่าว AI Aioga","description":"Xiaomi เปิดตัว Xiaomi-Robotics-1 ซึ่งรวมการฝึกล่วงหน้าแบบไม่ต้องมีร่างกายจำนวนมาก (UMI) กับข้อมูลหุ่นยนต์จริงเพียงเล็กน้อย หลังจากการฝึกล่วงหน้า ใช้ข้อมูลร่องรอย UMI 100,000 ชั่วโม...","url":"https://www.aioga.com/th/news/cmrsydl3b056vbitl77uxf5rp/","contentTranslated":true,"sourceHash":"88f9ed5cf2146ffd","translatedAt":"2026-07-22T19:29:43.266Z"},"pl":{"title":"Xiaomi wprowadza model strategiczny dla robotów Xiaomi-Robotics-1","summary":"Xiaomi wprowadziło Xiaomi-Robotics-1, które łączy wstępne szkolenie na dużą skalę bez ciała (UMI) z treningiem na niewielkiej ilości rzeczywistych danych z robotów. Wstępne szkolenie wykorzystuje 100 000 godzin danych trajektorii UMI obejmujących ponad 1700 scen, a dalsze szkolenie wprowadza ponad 7200 godzin rzeczywistych danych z domu. Model osiągnął SOTA na czterech benchmarkach symulacyjnych, a nowe zadania osiągają średnio 75% skuteczności po mniej niż 10 godzinach demonstracji.","category":"模型更新","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"Xiaomi wprowadza model strategiczny dla robotów Xiaomi-Robotics-1 - Aioga Wiadomości AI","description":"Xiaomi wprowadziło Xiaomi-Robotics-1, które łączy wstępne szkolenie na dużą skalę bez ciała (UMI) z treningiem na niewielkiej ilości rzeczywistych danych z robotów. Wstępne szkolen...","url":"https://www.aioga.com/pl/news/cmrsydl3b056vbitl77uxf5rp/","contentTranslated":true,"sourceHash":"88f9ed5cf2146ffd","translatedAt":"2026-07-22T19:30:33.654Z"}}}}