Core Insights - The article discusses the MiVLA model, which addresses the challenges of "data scarcity" and "generalization weakness" in the field of robot vision-language-action (VLA) models by utilizing a novel "human-robot mutual imitation pre-training" approach, allowing for effective training without real robot data [2][19] - MiVLA combines simulated robot data and human video data to achieve superior generalization capabilities, providing a low-cost and scalable path for general robot policy learning [2][19] Summary by Sections Need for Reconstructing VLA Pre-training Paradigm - Current VLA training faces dual challenges: reliance on real robot data is limited by high costs and limited scene coverage, while single-modal approaches suffer from "modal gaps" [3] - Effective VLA pre-training requires a unified approach that balances data scale, behavioral fidelity, and cross-modal adaptation [3] MiVLA's Design and Features - MiVLA's core design is based on aligning human and robot action spaces through mutual imitation pre-training, merging the diversity of simulated robot data with the fidelity of human video data [5] - Key features include: - Bidirectional human-robot action space mapping to overcome morphological differences [7] - Mutual imitation pre-training that leverages dual-source data advantages [8] - A diffusion transformer architecture to support continuous robot control [8] - Lightweight and efficient training for scalable deployment [8] Experimental Validation and Results - MiVLA was tested in both simulated and real robot environments, demonstrating significant performance improvements over baseline models [9][11] - In simulated tasks, MiVLA outperformed baseline models in 20 representative tasks, achieving an average success rate of 69% in easy mode and 66% in hard mode [10] - In real robot tasks, MiVLA matched the performance of large-scale real data pre-trained models using only medium-scale mixed data [11] Generalization Capability - MiVLA exhibited strong adaptability across different scenes, objects, and positions, achieving an average generalization success rate of 54% with only 20 demonstration data points [17][18] - The model's ability to handle unknown robot forms and complex tasks was validated through various experimental setups [11][14] Conclusion and Future Directions - MiVLA demonstrates that human-robot mutual imitation is key to overcoming data bottlenecks, allowing for the construction of a more generalized VLA model without real robot data [18] - Future improvements will focus on enhancing performance in extreme out-of-distribution scenarios, integrating multimodal information, and expanding data coverage [18]

Core Viewpoint - The article discusses the debate on whether embodied intelligence should be viewed as an application or as an independent foundational model, asserting that it is a foundational model specifically designed for the physical world, parallel to language and multimodal models [6][12][60]. Group 1: Differences Between Physical and Virtual Worlds - There is a fundamental difference between the physical world, characterized by randomness and continuous processes, and the virtual world, which is highly reproducible and low in randomness [2][10]. - Existing models based on language and visual modalities are inadequate for accurately representing the complexities and randomness of physical interactions [16][22]. Group 2: Need for a Separate Foundational Model - A separate foundational model for embodied intelligence is necessary due to the unique characteristics of the physical world, which often leads to unpredictable outcomes even under identical conditions [10][11]. - The current architectures and training methods struggle to capture the high randomness present in physical events, necessitating a new approach to model design [12][20]. Group 3: Future of Multimodal Models - Shifting the perspective to view embodied intelligence as an independent foundational model can lead to significant changes in model architecture and data utilization [9][23]. - The learning and perception processes in the physical world differ fundamentally from those in the virtual world, suggesting that future multimodal models should incorporate these differences [24][29]. Group 4: Scaling Laws and Data Utilization - The article emphasizes the importance of scaling laws in the development of large models, particularly in the context of robotics, where data acquisition and utilization are critical [46][51]. - A phased approach to training, utilizing both pre-training and post-training data, is recommended to enhance model performance [48][52]. Group 5: Hardware and AI Integration - The integration of AI in defining hardware is crucial for the development of embodied intelligence, advocating for a simultaneous evolution of both software and hardware [53][54]. - The potential for embodied intelligence to drive exponential growth in resources and capabilities is highlighted, suggesting a transformative impact on the future of artificial general intelligence (AGI) [59][60].

Group 1 - The core viewpoint of the article emphasizes the growth and development in the embodied intelligence sector, highlighting increased financing, production trials, and innovative product designs [2][3][4] - In financing, apart from a few star companies, the number of component companies has increased, and their financing amounts have grown [2] - In production, several companies are beginning pilot projects, with many startups seeking funding backed by orders, while leading humanoid robot companies are exploring industrial-grade product deployment [2] Group 2 - In product design, mechanical arm products are gradually converging, while innovations in structure and size continue in mobile operations and humanoid robots, with companies focusing on cost reduction and supply chain management [2] - The deployment of robots is advancing, with companies like Digua Robotics launching the S600 to support edge-side deployment, and Thor applying its technology in humanoid robots and mobile operations [4] - The computational power of over 2000T is becoming a reference configuration in the industry [4] Group 3 - The community is actively planning research reports and welcomes newcomers interested in the embodied intelligence field, having established various sharing platforms over the past year [7] - The community offers continuous live sharing sessions, roundtable forums, and a comprehensive technical roadmap for beginners [8][13] - It provides valuable industry systems and project proposals for those already engaged in related research [15][16] Group 4 - The community has established a job referral mechanism with multiple embodied companies, facilitating connections between job seekers and employers [18] - Members can access exclusive learning videos and documents, enhancing the learning experience [23] - The community has compiled a wealth of resources, including open-source projects, datasets, and technical learning routes, to support both newcomers and advanced learners [19][30]

负责公众号、小红书、社群的运营，提升粉丝粘性和关注度。我们希望您有一定的运营能力，对自媒体平台的 玩法有一定认识。 编辑岗位 咨询我们 负责日常公众号平台的内容创作、编辑，我们希望您具备一定的专业基础，在知乎、公众号等平台上具有内容 创作经验。 点击下方 卡片 ，关注" 具身智能 之心 "公众号 销售岗位 具身智能之心是具身领域的优秀技术创作平台，为行业输出了大量的前沿技术、课程、行业概况、融资、产 品、政策等内容。 负责平台课程、硬件等产品的销售推广。我们希望您具备一定的销售基础，对具身用户需求与市场有一定的了 解。 现平台正处于上升期，因业务需求，面向全体粉丝招募编辑、运营、销售岗位，和我们一起继续为领域创造价 值，全职+实习哦（实习除编辑岗位均需线下哦～） 运营岗位 如果您有兴趣和我们一起成长，欢迎添加峰哥微信oooops-life ...

Core Insights - The article discusses the current state of robot learning as of December 2025, emphasizing that most systems rely on behavior cloning (BC) and the challenges associated with it [5][40][39] - It highlights the importance of data collection from human demonstrations and the limitations of existing methods in achieving robust performance in real-world applications [6][10][12] Group 1: Behavior Cloning and Its Challenges - As of December 2025, all robot learning systems primarily utilize behavior cloning, where human demonstrations are used to train models to mimic actions [5] - The data for behavior cloning comes from human demonstrations and various other sources, but the need for extensive data collection poses significant challenges [7][10] - The limitations of behavior cloning include the inability to generalize well to out-of-distribution (OOD) states, leading to performance degradation in real-world scenarios [16][23][40] Group 2: Data Collection Methods - Data collection methods include using human operators with smart demo gloves and video platforms to gather diverse task execution data [11][13] - The challenges in data collection include ensuring the data is representative of the tasks and the need for extensive training for operators to provide usable data [9][10] - The article emphasizes the importance of high-quality data for training models and the difficulties in achieving this at scale [10][19] Group 3: Future Directions in Robot Learning - The article predicts that within two years, video model backbones will replace current VLA methods, and within ten years, world models will effectively simulate general open-world interactions [73] - It suggests that traditional simulation and game engines will serve as data generators for world models, emphasizing the continued importance of expert demonstration data [73] - The need for robust Q/V functions that can operate effectively in OOD states is highlighted as a critical area for future research [72]

Core Viewpoint - The article discusses the application of Reinforcement Learning (RL) in enhancing Text-to-3D generation, exploring its effectiveness and challenges in this complex domain [4][5]. Group 1: Research Background - A collaborative research effort involving multiple universities aims to investigate the potential of RL in improving 3D generation processes [4]. - The study focuses on whether RL can enhance the reasoning and generation capabilities of 3D autoregressive models, building on its success in large language models (LLMs) and 2D image generation [5]. Group 2: Challenges in 3D Generation - Key challenges identified include designing rewards that capture semantic alignment, geometric consistency, and visual quality [6]. - Existing RL algorithms may not be suitable for autoregressive 3D generation, and there is a lack of benchmarks specifically assessing "3D reasoning capabilities" [6]. Group 3: Reward Design Layer - The research found that aligning with human preference signals is crucial for improving overall 3D quality, while specialized reward models often outperform large multimodal models [10]. - The study indicates that token-level strategies in RL are more effective than sequence-level operations in 3D autoregressive generation [11]. Group 4: Benchmark Layer - The MME-3DR benchmark was developed to evaluate 3D reasoning, focusing on maintaining consistency and interpretability under challenging constraints [15]. - RL training significantly improved performance across various tasks, particularly in mechanical structures and non-rigid biological entities [16]. Group 5: RL Paradigm Layer - The research proposes a hierarchical RL paradigm (Hi-GRPO) that treats 3D generation as a coarse-to-fine process, enhancing the model's implicit 3D reasoning capabilities [18][19]. - The findings highlight the importance of respecting structural priors in the design of reward models for effective training [20]. Group 6: Performance Insights - The study reveals that while RL can enhance model performance, challenges remain in handling complex geometries and rare concepts, indicating limitations in current 3D RL capabilities [22].

Core Viewpoint - The article discusses the rapid development and challenges of the VLA (Whole Body Visual Learning Algorithm) in the field of embodied intelligence, highlighting the importance of real data collection and the difficulties faced by newcomers in the field [2][3][4]. Group 1: VLA Development and Challenges - The VLA algorithm is experiencing explosive growth, with various frameworks and tools, such as reinforcement learning (RL), enhancing its performance [2]. - Data collection methods are diversifying, with millions of open-source data becoming available, indicating a potential for industrialization [2]. - Many practitioners express frustration with the challenges of tuning VLA models and the complexities of data collection, particularly for those new to the field [3][5]. Group 2: Data Collection and Training - Data collection methods for VLA primarily include imitation learning and reinforcement learning, with a focus on remote operation and VR for mechanical arms [13]. - Simulation and real-to-sim-to-real (real2sim2real) techniques are crucial for training VLA models, especially when real data is insufficient [14]. - Training techniques are critical, with many practitioners struggling to achieve good results due to the complexity of models like π0 and π0.5, which require high attention to detail [14][10]. Group 3: Model Deployment - After training, VLA models require optimization to reduce their parameter size for deployment, which is essential for edge computing applications [15]. - Techniques such as quantization and distillation are necessary to maintain performance while minimizing model size [15]. Group 4: Educational Initiatives - The article introduces a practical course aimed at helping individuals learn VLA effectively, covering hardware, data collection, algorithm deployment, and real-world experiments [17][20]. - The course is designed to save time and reduce the learning curve for newcomers, providing practical experience that can enhance resumes [18][31].

编辑丨 RoboX 点击下方 卡片 ，关注" 具身智能之心 "公众号 >> 点击进入→ 具身 智能之心 技术交流群 更多干货，欢迎加入国内首个具身智能全栈学习社区： 具身智能之心知识星球（戳我） ，这里包含所有你想要的! 近年来，具身智能提出了新的仿真数据需求——既要求3D场景不仅要具有照片级的真实感，还要求场景中的每个实例都能在物理层面上进行交 互，以支持在仿真环境中训练机器人策略。 在这其中，核心的桌面场景（Tabletops）是此类环境的「最后一步」，也是大多数精细交互和复杂机器人操作任务的基础舞台。 因此，自动化、大规模地生成高保真、可交互的桌面场景，对于推进具身操作策略学习至关重要。 在此背景下，地瓜机器人联合中国科学院大学、地平线、 中科院自动化所等发布了今年的关键研发成果—— TabletopGen ：一个统一的、无 需训练的桌面场景生成框架。 3D仿仍存严重不足 据介绍，现有的仿真方法仍存在严重不足： 1、文本驱动方法的局限性 ： 例如Holodeck[1]，它利用大语言模型（LLM）直接生成 3D 布局，或者通过生成场景图或空间约束，再进行 布局可行性的优化。 然而，这两类路径通常都只是从固 ...

Core Viewpoint - The article discusses the launch of innovative technologies by DaXiao Robotics, including the ACE embodiment research paradigm, the open-source Kairos 3.0 world model, and the embodiment super brain module A1, aimed at creating a self-controlled, open, and win-win industrial ecosystem for embodied intelligence [1][3][25]. Group 1: ACE Research Paradigm and World Model - DaXiao Robotics has introduced the ACE embodiment research paradigm, which fundamentally innovates the research path for embodied intelligence by focusing on human-centric data collection and interaction with the physical world [12][14]. - The Kairos 3.0 world model enhances the value of real data, achieving a scale of over 100 million hours of data, which is crucial for the development of embodied intelligence [12][18]. - The ACE paradigm allows for the collection of over 10 million hours of data annually through environmental data collection techniques, which integrate various modalities such as visual, tactile, and auditory data [12][14]. Group 2: Technological Innovations and Applications - The embodiment super brain module A1 enables robots to operate in complex and dynamic environments without pre-collected high-precision maps, showcasing robust path generation capabilities [25][27]. - The A1 module can interpret natural language commands and execute tasks with high precision, making it suitable for various applications across multiple industries [27][29]. - DaXiao Robotics has established partnerships with leading chip manufacturers and cloud service providers to enhance the performance of its technologies and facilitate the rapid development of customized embodied intelligence products [23][32][34]. Group 3: Industry Collaboration and Ecosystem Development - The company emphasizes the importance of ecosystem collaboration, working with various partners across the supply chain, including hardware, chip, and cloud service providers, to create a comprehensive and self-controlled ecosystem for embodied intelligence [30][31]. - DaXiao Robotics aims to lead the scale development of China's embodied intelligence industry by continuously innovating technologies and fostering collaborative efforts with industry partners [31][33].

Financing - In the second half of the year, apart from some star companies, the financing amount for core component companies has increased, and the number of companies has also grown [2] Mass Production - Several companies have begun pilot projects, with many startups coming in with orders for financing. Leading humanoid robots are starting to explore industrial-grade product deployment [2] Product Design - Mechanical arm products are gradually converging, while innovations in structure and size continue in mobile operations and humanoid designs. Companies are also focusing on cost reduction, with supply chain management capabilities significantly influencing future competitiveness. Leading companies in the field are actively investing in component suppliers [2] Model Generalization - The optimization approach based on Reinforcement Learning (RL) is enhancing the generalization capabilities of models. Related toolkits are becoming more refined, making real machine deployment increasingly convenient [3] Deployment - DiGua Robotics has launched the S600 to assist with edge-side deployment. Thor is beginning to be applied in humanoid robots and mobile operations, with computing power above 2000T becoming a reference configuration [4] Community Development - The community is actively planning research reports and welcomes newcomers interested in entering or advancing in the field of embodiment. Over the past year, the community has completed various segments including technical route sharing, live broadcasts, Q&A, job seeking, and competitions [6] Technical Sharing - The community has prepared numerous roundtable forums and live broadcasts covering various topics from embodiment, data, to algorithms, gradually sharing what is happening in the embodiment industry and the problems that remain to be solved [8] Technical Roadmap - For beginners, the community has organized many technical stacks and routes to facilitate entry into the field [11] Industry and Project Solutions - For those already engaged in related research, the community provides valuable industry systems and project solutions [14] Job Referral and Career Support - The community has established a job referral mechanism with several embodiment companies, allowing members to submit their resumes directly to desired companies [16] Resource Compilation - The community has compiled a comprehensive list of open-source projects, datasets related to embodiment intelligence, and mainstream simulation platforms to assist members in their learning and research [17]