2025年，随便刷一刷汽车新闻，就会发现大家讨论的焦点变了。以前，公众可能更关注发动机、百公里加速时间；而现在，电动隐藏式车门把手的是是 非非，动力电池新国标到底有多严，甚至电动汽车"反向送电"如何收益，这些关于具体部件、标准和模式的话题，频频冲上新闻热搜。 新的话题标签，其实都指向同一个大变化：汽车的核心竞争力正从传统的机械性能，转向是否足够"聪明"、足够安全，能否融入我们的生活与能源网络等方 面。细微和琐碎的零部件，开始成为搅动整个汽车产业变革的"风暴眼"。 本刊编辑部整理了2025年受到关注的8个零部件领域热词，希望能从这些具体的"小切口"入手，为读者呈现一幅汽车产业链正在发生的大变化。 大模型 2025年，AI大模型持续升温。VLA（视觉-语言-动作）模型、VLM（视觉语言大模型）与世界模型正与汽车产业形成深度协同，重塑智能汽车的感知、决策 与交互体系。同时，算力芯片的激烈竞争、电子电气架构向中央集中式的演进，以及"软件定义汽车"商业模式的成熟，共同为大模型上车铺平道路。 VLM的研发与应用，主要聚焦于智能座舱的升级，即通过整合OMS（车内乘员监测摄像头）、语音传感器等硬件数据，构建个性化用户服务体系 ...

大家早上壕！ 先看海外要闻： 道指盘中首次站上49000点，随后涨幅回落但依旧创历史新高，小盘股指涨幅居前。美股收盘标普500 涨幅0.64%，道指涨幅1.23%，纳指涨幅0.69%。 特斯拉涨超3%，亚马逊涨逾2%，苹果跌超1%。雪佛龙上涨超5%，特朗普表示大型美国石油公司将 投资委内瑞拉。 纳斯达克金龙中国指数收涨0.49%，中概股索威尔收涨超53%，脑再生科技涨超31%，老虎证券涨超 6%，再鼎医药涨超4%。 避险情绪下，美债收益率普跌，疲软的ISM数据公布后，10年期收益率进一步下行、走低4.3个基点。 美元下跌，回吐了因短线避险情绪带来的涨幅。比特币大涨，重返9.5万美元附近。以太坊一个月来 首次站上3200美元关口。 贵金属大涨，现货黄金涨超2.5%，现货白银大涨5%，纽约银日内一度暴涨近10%。伦铜首次升至 13000美元。原油先跌后涨，WTI原油较日低涨超3.6%。 英伟达在CES发布新一代Rubin平台，推理成本较Blackwell降10倍，拟下半年发货。英伟达"内驱"无 人驾驶汽车将至，发布首个链式思维推理VLA模型。 背靠OpenAI的机器人初创1X亮相CES展示家务机器人，售价2万美元 ...

Investment Rating - The industry investment rating is "Positive" [2] Core Insights - The report highlights the successful launch of China's first batch of L3 autonomous vehicles in Chongqing, with 46 vehicles now operational [22] - Geely Automobile has completed the privatization of its subsidiary, Zeekr, which is now a wholly-owned subsidiary [22] - The report emphasizes the gradual relaxation of intelligent driving policies, which is expected to drive growth for related companies [3] Summary by Sections Market Performance - The A-share automotive sector outperformed the market, with a weekly increase of 2.74%, compared to a 1.95% rise in the CSI 300 index [3][9] - The passenger vehicle sector saw a 3.26% increase, led by BYD and Haima Automobile [3][17] - The commercial vehicle sector experienced a slight decline of 0.02%, while the automotive parts sector rose by 3.32% [3][20][21] Key Industry News - The first L3 autonomous vehicles have been deployed in Chongqing, focusing on complex traffic conditions [22] - Geely's Zeekr has been privatized and delisted from the NYSE [22] - Beijing has issued the first special license plates for L3 autonomous vehicles [22] - Shenzhou Car Rental has initiated autonomous driving tourism tests in Hainan [22] - The VLA model by Yuanrong Qixing has entered mass production, marking a significant technological advancement [22] - Baidu and Uber are collaborating to test autonomous ride-hailing services in the UK, expected to launch by the end of next year [22] Upstream Data Tracking - The report includes tracking of key material prices such as steel, aluminum, natural rubber, and lithium carbonate, which are crucial for the automotive industry [24][26]

Domestic News - The Ministry of Transport aims to accelerate the development of smart logistics, low-carbon economy, and digital transportation industries, focusing on integrating AI and new technologies into transportation systems [7] - The State-owned Assets Supervision and Administration Commission emphasizes the need for state-owned enterprises to continue focusing on key areas such as new energy, new energy vehicles, and advanced materials, while promoting strategic mergers and acquisitions [8] - By 2026, over 10,000 charging stations will be built in highway service areas across China, with at least 25% being high-power chargers [9] - Beijing has issued the first L3-level autonomous driving vehicle license plates, marking a significant milestone in the country's autonomous vehicle development [10] - SAIC-GM-Wuling and Tsinghua University have established a joint research center for automotive AI, focusing on smart driving technologies [11] - CATL has launched the longest battery swap route in China, covering 1,250 kilometers, enhancing the electric vehicle infrastructure [12] - Shenzhou Car Rental has initiated autonomous driving tourism tests in Hainan, creating the world's first L4-level autonomous driving tourist route [13] - The Yuanrong Qixing VLA model has been mass-produced, representing a significant advancement in automotive technology [14] International News - In November, European car sales increased by 2.4% year-on-year, reaching 1.08 million units [16] - Tesla has signed a contract with Matrix Renewables for a 1GWh Megapack energy storage project in Scotland [17] - Tata Motors plans to launch five new electric vehicle models in India by March 2030, aiming to capture a 50% market share in the electric vehicle sector [18] - Samsung SDI is collaborating with KG Mobility to develop advanced battery pack technology for electric vehicles [19] Commercial Vehicles - Foton's Pro Series low-entry trucks have been unveiled, focusing on high efficiency and user-friendly design for urban sanitation [20] - Jiefang Qingqi has launched its high-cold testing for 2025-2026, with over 30 vehicles participating in the tests, marking a significant scale for the company [21] - The Jianghuai 1 Card Kunpeng ET9 has been launched in South China, achieving a remarkable range of 3.7 kilometers per kilowatt-hour under highway conditions [22] - Jiangqi Group has officially launched its light and heavy truck products in the Brazilian market, marking a key milestone in its South American strategy [23]

Core Insights - The article discusses the rapid development of Vision-Language-Action (VLA) models in embodied intelligence, highlighting their unprecedented generalization capabilities but also addressing the critical issue of instability during the reasoning phase [2][3][4]. - A novel framework named TACO (Test-time Anti-exploration via pseudo-Counts) is introduced to tackle the reasoning instability in VLA models, providing a solid theoretical foundation and practical solutions [2][8]. Group 1: VLA Model Challenges - VLA models, despite their impressive average performance, exhibit extreme sensitivity to initial noise during inference, leading to success rates that can fluctuate between 0% and 80% for the same model [4][6]. - The instability is attributed to two main factors: the retention of redundant action patterns from diverse pre-training data and the multimodal nature of fine-tuning datasets, which may include suboptimal strategies [7][6]. Group 2: TACO Framework - TACO draws inspiration from the "anti-exploration" principle in offline reinforcement learning, aiming to constrain generated actions to successful patterns within the fine-tuning dataset [9][11]. - The framework includes three key components: a Coupled Pseudo-Count Estimator that utilizes the VLA model's internal representation, ensuring efficient validation without additional training [11][12]. Group 3: Implementation and Results - TACO employs a two-stage reasoning process: generating diverse action candidates and validating them through pseudo-counts, which are calculated using a trained CFN [17][18]. - The implementation of a Shared Observation Key-Value Cache significantly reduces computational costs, allowing for efficient real-time operation with minimal latency [20][21]. Group 4: Experimental Validation - Comprehensive evaluations across multiple simulated benchmarks and a dual-arm robot platform demonstrate TACO's effectiveness, with average success rates improving by 16% in real-world tasks [22][32]. - Specific tasks, such as "organizing paper and pens," showed a remarkable 25% increase in success rates, highlighting TACO's ability to filter out suboptimal behaviors [32][33]. Group 5: Future Directions - TACO not only addresses practical challenges but also opens new perspectives for VLA research, suggesting potential expansions into more complex multi-task scenarios and integration with world models for enhanced long-term planning capabilities [35].

目前，元戎启行的工程化能力已趋成熟，为VLA模型的量产上车提供坚实基础。最新数据显示，元戎启行已交付20万辆搭 载城市NOA（领航辅助驾驶）的量产车型，合作车型覆盖SUV、MPV、越野车等多个品类。10月份，元戎启行在辅助驾驶城 市NOA第三方供应商市场的单月市占率接近40%。 通过魏牌蓝山智能进阶版的正式发布，元戎启行VLA模型为用户带来更安心、更可靠的辅助驾驶体验。未来，伴随着元戎 启行辅助驾驶方案的持续上车，VLA模型将在更广泛的使用场景中发挥关键价值，持续推动辅助驾驶向更高水平的使用体验迈 进。 展望2026年，元戎启行将推进Robotaxi业务和RoadAGI业务。其中，量产车辅助驾驶业务将继续扩大合作车型与客户体 系，力争在2026年实现累计交付突破一百万辆的目标。 （编辑 张明富） 本报讯 （记者王镜茹）近日，魏牌全新蓝山智能进阶版正式上市，成为全球首款搭载深圳元戎启行科技有限公司（以下简 称"元戎启行"）VLA（Vision-Language-Action）模型的量产车型。这不仅意味着VLA模型完成从技术研发到量产上车的闭环验 证，更标志着元戎启行的商业化落地能力迈入全新阶段。 元戎启行CEO ...

Core Viewpoint - The article emphasizes the increasing demand for VLA (Variable Latent Action) algorithms in the industry, highlighting the challenges associated with data collection and model training, which are critical for successful implementation in real-world applications [1][2][3]. Group 1: VLA Algorithm Demand and Challenges - There is a significant demand for VLA algorithms, as evidenced by numerous job postings and the increasing number of related research papers [1]. - Many practitioners express frustration over the difficulties in tuning VLA algorithms and the complexities involved in data collection [2]. - The reliance on real machine data for effective VLA model training poses challenges, as the data collected often proves to be inadequate for practical applications [3][8]. 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 technologies [10]. - Effective data collection and ensuring high-quality data are crucial, particularly in the context of real-to-sim-to-real (real2sim2real) methodologies [10]. - Training VLA models typically requires simulation debugging, especially when real machine data is insufficient, with frameworks like Mujoco and Isaac Gym being essential for this process [11]. Group 3: Model Deployment and Optimization - After training, VLA models often require optimization techniques such as quantization and distillation to reduce parameter size while maintaining performance [12]. - The deployment of VLA models on edge devices presents challenges due to their large parameter sizes, necessitating lightweight operations [12]. - The article discusses the importance of fine-tuning models and the various tricks involved in training complex models like π0 and π0.5, which require significant expertise [11][8]. Group 4: Educational Initiatives - The article introduces a practical course aimed at helping individuals learn about VLA, covering topics such as hardware, data collection, algorithm training, and model deployment [13][17]. - The course is designed to address the rapid advancements in VLA technology and aims to equip participants with hands-on experience and knowledge [13][18]. - It includes a comprehensive curriculum that spans various aspects of VLA, from foundational concepts to advanced deployment techniques [19][20][21].

人民财讯12月23日电，12月18日，优必选旗下的智慧物流子公司UQI优奇与字节跳动旗下云和AI服务平 台火山引擎在FORCE原动力大会正式签署合作协议。根据协议，双方将在多模态大模型、VLA模型， 豆包生态交互、AI云原生基础设施等方面展开深入合作，为人形机器人、无人物流车、工业移动机器 人等实体机器人在工业及物流垂直领域的AI应用提供技术动能。围绕本次合作，双方将从算法联合研 究、产品及工程开发合作、数据中心及工具链共建三个领域展开深度协同，并形成落地路径。 ...

Core Insights - The VLA (Visual-Language-Action) model faces challenges in training due to the scarcity and complexity of physical world data, which is essential for enhancing its capabilities [2] - The shift from a machine-centered to a human-centered AI research paradigm is necessary for training embodied intelligence models [2] Group 1: ACE Paradigm - The ACE (Action-Centric Embodiment) paradigm focuses on human interaction with the physical world as the starting point for research, utilizing environmental data collection as a core engine [3] - The ACE framework integrates first-person and third-person video, haptic information, motion trajectories, and voice data to create a physics-based 3D asset library [3] Group 2: Open Source Model - The "Awakening World Model 3.0" has been released as an open-source and commercially applicable model, aiming to unify understanding across different domains by integrating physical laws, human behavior, and real machine actions [5] - The platform supports 11 major categories and 54 subcategories, covering 115 types of embodied scenarios, allowing developers to quickly generate visual task simulations with simple commands [5] Group 3: Industry Collaboration - The company collaborates with various robotics firms to integrate the ACE paradigm and world model with hardware solutions tailored for different scenarios [6] - Partnerships with domestic chip manufacturers enhance the computational capabilities of the Awakening World Model 3.0 [6] Group 4: Market Applications - In the short term, the focus will be on deploying quadruped robots with autonomous navigation capabilities in security and inspection sectors [6] - The medium-term strategy includes addressing labor-intensive logistics scenarios, while long-term goals involve exploring applications in home environments, contingent on resolving safety and liability issues [6] Group 5: Industry Trends - The ACE paradigm aligns with trends observed in companies like Tesla and Figure AI, indicating a significant opportunity in the sector over the next one to two years [7]

Core Viewpoint - The article discusses the rising interest and investment in the field of embodied intelligence, particularly in humanoid robots, highlighting the shift in investor focus and the challenges faced by startups in this sector [5][6][13]. Investment Trends - In 2023, there has been a significant influx of venture capital into the embodied intelligence sector, with estimates suggesting over 100 active investment firms and early-stage funding exceeding $10 billion in China [6]. - Investors are particularly interested in startups led by individuals with backgrounds in intelligent driving, as they bring valuable experience in productization and operational expertise [6][7]. Entrepreneurial Landscape - The article identifies a new wave of entrepreneurs in the embodied intelligence space, many of whom have transitioned from the intelligent driving industry, including notable figures from companies like Huawei, Xpeng, and Baidu [7][8]. - The "Berkeley Four," a group of entrepreneurs from the University of California, Berkeley, have gained attention for their contributions to the field, reflecting a shift in investor preferences towards teams with practical experience [7]. Technological Challenges - The transition from intelligent driving to embodied intelligence involves overcoming significant technical hurdles, including the need for high-quality interaction data and the development of robust algorithms capable of generalizing across various tasks [12][10]. - Current embodied robots face challenges in cost-effectiveness, with prices for certain models around 600,000 yuan (approximately $90,000), which may decrease to 350,000-400,000 yuan (about $50,000-$60,000) by 2027, but this does not account for maintenance and operational costs [12]. Market Sentiment - There is a growing skepticism in the secondary market regarding the sustainability of investments in embodied intelligence, with some analysts suggesting that the best opportunities may have already passed [13]. - The article notes that the number of humanoid robot companies in China has surpassed 150, raising concerns about market saturation and the potential for a bubble in the sector [13]. Investment Logic - Investors are prioritizing projects that focus on the core components of embodied intelligence, including decision-making models, control systems, and the physical robots themselves, while also being cautious of the high similarity in pitches from various startups [14][15].