据悉，商汤绝影的世界模型研发负责人离职。据了解其负责商汤绝影云端技术体系建设，也是绝影的生 成式智驾方案R-UniAD的研发负责人。 以下文章来源于红色星际 ，作者红色星际科技 红色星际 . 让更多人，更深入地了解自动驾驶行业！ 来源 | 红色星际 点击下方 卡片 ，关注" 自动驾驶之心 "公众号 戳我-> 领取 自动驾驶近15个 方向 学习 路线 >>点击进入→ 自动驾驶之心 『行业第一线』技术交流群 本文只做学术分享，如有侵权，联系删文 传闻该负责人离职之后有可能去创业。 25年对于商汤绝影来说仍旧会是充满挑战的一年。 在中阶上，商汤绝影在广汽传祺上量产交付了基于J6M的中阶方案。不过，今年中阶市场将迎来升级迭 代，也就是基于J6M做出来城区NOA，从目前的高速NOA升级为全域NOA，这是今年中阶市场最大的变 化。 一些头部公司基于高阶方案的蒸馏剪裁推出了轻量版的城区NOA，下放到100 TOPS算力左右的芯片上， 而且已经给主机厂客户做推广演示。 这次算法方案的升级迭代也会改变中阶市场的格局。跟得上的玩家，将会继续留在牌桌上，而跟不上 的，就有可能被挤下来。 这是行业非常残酷的一面，算法方案每年一次迭代 ...

Core Insights - The article presents MMTL-UniAD, a unified framework for multimodal and multi-task learning in assistive driving perception, which aims to enhance the performance of advanced driver-assistance systems (ADAS) by simultaneously recognizing driver behavior, emotions, traffic environment, and vehicle actions [1][5][26]. Group 1: Introduction and Background - Advanced driver-assistance systems (ADAS) have significantly improved driving safety over the past decade, yet approximately 1.35 million people die in traffic accidents annually, with over 65% of these incidents linked to abnormal driver psychological or physiological states [3]. - Current research often focuses on single tasks, such as driver behavior or emotion recognition, neglecting the inherent connections between these tasks, which limits the potential for cross-task learning [4][3]. Group 2: Framework and Methodology - MMTL-UniAD employs a multimodal approach to achieve synchronized recognition of driver behavior, emotions, traffic environment, and vehicle actions, addressing the challenge of negative transfer in multi-task learning [5][26]. - The framework incorporates two core components: a multi-axis region attention network (MARNet) and a dual-branch multimodal embedding module, which effectively extract task-shared and task-specific features [5][26]. Group 3: Experimental Results - MMTL-UniAD outperforms existing state-of-the-art methods across multiple tasks, achieving performance improvements of 4.10% to 12.09% in the mAcc metric on the AIDE dataset [18][26]. - The framework demonstrates superior accuracy in driver behavior recognition and vehicle behavior recognition, with increases of 4.64% and 3.62%, respectively [18][26]. Group 4: Ablation Studies - Ablation experiments indicate that joint training of driver state tasks and traffic environment tasks enhances feature sharing, significantly improving task recognition accuracy [22][26]. - The results confirm that the interdependence of tasks in MMTL-UniAD contributes to overall performance and generalization capabilities [22][26].

Core Viewpoint - The article discusses the challenges and opportunities faced by SenseTime's autonomous driving division, particularly focusing on the competitive landscape and the importance of technological advancements in the industry. Group 1: Company Developments - The head of the world model development for SenseTime's autonomous driving division has left the company, which raises concerns about the future of their cloud technology system and the R-UniAD generative driving solution [2][3]. - SenseTime's autonomous driving division has successfully delivered a mid-tier solution based on the J6M model to GAC Trumpchi, but the mid-tier market is expected to undergo significant upgrades this year [4]. Group 2: Market Dynamics - The mid-tier market will see a shift from highway-based NOA (Navigation on Autopilot) to full urban NOA, which represents a major change in the competitive landscape [4]. - Leading companies are introducing lightweight urban NOA solutions based on high-tier algorithms, targeting chips with around 100 TOPS computing power, which are already being demonstrated to OEM clients [4]. Group 3: High-Tier Strategy - The key focus for SenseTime this year is the one-stage end-to-end solution, which has shown impressive performance and is a requirement for high-tier project tenders from OEMs [5]. - Collaborations with Dongfeng Motor aim for mass production and delivery of the UniAD one-stage end-to-end solution by Q4 2025, marking a critical opportunity for SenseTime to establish a foothold in the high-tier market [5][6]. Group 4: Competitive Landscape - SenseTime's ability to deliver a benchmark project in the high-tier segment is crucial for gaining credibility with OEMs and securing additional projects [6][7]. - The current window of opportunity for SenseTime in the high-tier market is limited, as many models capable of supporting high-tier software and hardware costs are being released this year [6][8].

Core Insights - The 2025 Shanghai Auto Show highlights the transformation of the automotive industry, showcasing innovations in the parts sector driven by electrification, intelligence, and globalization [2][3] - The event reflects the automotive supply chain's cross-domain collaboration and the industry's diverse and global landscape [2][3] - The shift from hardware-dominated to software-defined vehicles marks a fundamental change in industry logic, with software becoming a key factor in defining automotive product capabilities [3][4] Industry Trends - The dual drive of electrification and intelligence is the core theme of the 2025 Shanghai Auto Show, with a significant increase in the number of exhibitors showcasing new technologies and products [3][5] - The automotive software market is projected to reach $55 billion by 2025, with over-the-air upgrades, smart cockpit interactions, and autonomous driving algorithms as the main growth areas [6] - AI technology is increasingly integrated into vehicles, enhancing their capabilities and transforming them into mobile smart terminals [4][6] Company Strategies - Foreign parts manufacturers are shifting from unilateral technology output to deep ecological integration, focusing on localizing technology implementation [9][12] - Bosch aims to leverage software and technological innovations to empower smart mobility, with a projected 4% sales growth in China for 2024, reaching 116.6 billion yuan [9] - Continental Group plans to list its automotive subgroup AUMOVIO, emphasizing its commitment to the Chinese market and local innovation [10][11] Emerging Technologies - The integration of AI and automotive technology is reshaping the collaboration model among people, vehicles, roads, and clouds, positioning smart vehicles as essential components of a smart lifestyle [4][5] - New products such as the R-UniAD technology from SenseTime and the advanced chip platforms from Black Sesame are pushing the boundaries of autonomous driving capabilities [7][6] - The introduction of immersive smart cockpit solutions and multi-modal technologies is enhancing user experience and safety in vehicles [8][6] Market Dynamics - The presence of over 300 exhibitors, including major international tech companies, indicates a significant shift in the automotive market landscape, with parts manufacturers taking center stage [13][15] - The competition in the automotive industry is evolving from hardware breakthroughs to collaborative innovation across the entire supply chain [5][12] - The 2025 Shanghai Auto Show serves as a critical observation point for the ongoing transformation in the automotive value chain, with a notable shift towards electrification and intelligence [16]

这场技术革命已超越单一性能的比拼，转向全场景体验的整合。长安汽车的"天衡底盘"与奔驰的"驾控 超级大脑"将机械性能与数字算法深度融合，实现毫秒级动力响应与复杂路况预判。值得关注的是，中 国品牌以"全栈自研"重塑产业话语权：华为乾崑智驾ADS4系统与鸿蒙座舱HarmonySpace5的落地，标 志着"人—车—家"生态闭环的成熟；而商汤绝影的R-UniAD技术方案则通过世界模型与强化学习，推动 智驾系统从"被动应对"向"主动思考"进化。 汽车产业的边界正在消融。华为、地平线等科技企业与车企的深度耦合，使汽车演变为移动智能终端。 用户价值体系的重构更具深意。蔚来以"油电同质同价"策略打破能源形式藩篱；沃尔沃XC90的透明车 身展示硼钢闭环结构，将安全从技术参数强化为品牌基因。 ■刘钊 参展面积从1.5万平方米到36万平方米、参展商从73家到近1000家，历经40年的发展，上海车展目前已 成为全球规模最大的汽车行业展会。日前开幕的第二十一届上海国际汽车工业展览会（以下简称"2025 上海车展"）汇聚了来自26个国家和地区的近千家企业。中国汽车产业正以技术创新为引擎、以生态重 构为路径，向世界展示着从"市场跟随者"到" ...

" 更低成本的真实数据需求，更快迭代智驾系统。 " 作者丨田哲 编辑丨林觉民 4月23日，第二十一届上海国际车展上，商汤绝影发布了包括生成式智驾R-UniAD、4D世界模型"绝影开 悟2.0"、AI内核"绝影千机"在内的多个核心技术。 统观绝影发布的技术，一句话概括为：不是做"功能模型"，而是做"系统AI"。 过去几年，端到端技术被认为是智能驾驶实现路径中的"终极形态"，但大多数方案都卡在数据、性能、安 全三道门槛前。R-UniAD的思路，是在这三道门槛外，再造一层AI基础设施。 据介绍，R-UniAD结合世界模型和强化学习，以VLAR（视觉、语言、行动、强化学习）技术架构为核 心，支持仿真环境中进行闭环训练，让模型能在虚拟世界里"试错—强化—泛化"，最终形成系统策略能 力。 以车展现场展示的"施工占道场景"为例，R-UniAD先通过"绝影开悟"世界模型构建4D动态环境，再让模 型反复在这个世界里强化训练不同策略。最终结果是，模型不仅能精准避让，还能泛化到其它类似场景 中。 这种能力的意义在于，它不仅降低了数据成本，更提高了模型上限——在少数据场景里跑出更强泛化能 力，是当前行业少见的尝试。 如果说R-Uni ...

Core Viewpoint - The collaboration between SenseTime and Dongfeng Motor in the field of intelligent driving represents a deep integration of technology and industry, providing a practical model for the sector [1] Technological Breakthroughs - SenseTime showcased the R-UniAD solution based on the VLAR architecture, achieving near real-time interactive simulation and reducing the need for real data by two orders of magnitude [3] - The system demonstrated a perception accuracy of 98% compared to real data and improved simulation efficiency by five times over industry benchmarks [3] - The joint team has made significant progress in developing an end-to-end autonomous driving system, which features a "dual insurance" architecture combining a learning-based decision system with a rule-driven safety redundancy [3] Industry Collaboration - SenseTime has partnered with four car manufacturers, covering seven vehicle models, and plans to mass-produce an end-to-end solution based on the NVIDIA Thor platform by Q4 2025 [3] - The company has maintained a leading market share in intelligent cockpit software for five consecutive years, with cumulative deliveries exceeding 3.6 million units [3] Ecological Evolution - SenseTime introduced the first in-vehicle AI operating system, "Jueying Qianji," with a response latency of less than 300ms, supporting over 100 intelligent agents [6] - The upgraded "New Member" system showcases multimodal recognition and deep thinking capabilities, providing personalized services during real-time demonstrations [6] - The "Travel Medical" system integrates medical-grade health monitoring and connects with third-party medical services [8] Industry Observations - The recent regulatory adjustments are viewed as a turning point towards maturity rather than a pause in the intelligent transformation process [8] - The penetration rate of L2-level assisted driving reached 62% in 2024, with a 37% year-on-year increase in user complaints regarding functional safety [8] - The collaboration model between Dongfeng and SenseTime reflects a dual-track strategy of "independent control + open collaboration" in the intelligent transformation of traditional car manufacturers [8] Common Industry Challenges - The introduction of AI-native technologies like world models and reinforcement learning is breaking through the limitations of rule-driven systems [11] - There is an urgent need to establish compliant data collection systems and unified safety assessment standards within the industry [11] - The deep binding of AI companies and vehicle manufacturers may be a key pathway to address these common challenges [11]

" 商汤绝影将于今年4月上海车展发布R-UniAD端到端自动驾驶方 案，完成实车部署，并预计在年底交付。 " 作者丨李雨晨 雷峰网获悉，哪吒汽车产品研发中心副总经理、智能驾驶部高级总监王俊平已于今年2月加入商汤绝影。 再早之前，王俊平是百度智能驾驶团队的一员。 2024年，石建萍曾透露对外合作的进展。她表示，目前商汤合作的车企有30多家，包括广汽、比亚迪、 本田、蔚来等。在智能驾驶方面，解决方案已经上线昊铂，还有哪吒的超级轿跑车型。 2025年2月，商汤绝影CEO王晓刚发布了"与世界模型协同交互的端到端智能驾驶路线R-UniAD"，以绝 影量产智驾方案和"开悟"世界模型为核心，实车采集的真实数据和云端生成的仿真数据进行闭环流转。按 照计划，商汤绝影将于今年4月上海车展发布R-UniAD端到端自动驾驶方案，完成实车部署，并预计在年 底交付。 （关于商汤绝影的智能化研发进展及人事调整，感兴趣的读者可添加作者微信Gru1993交流） // 近期热门文章 哪吒和商汤在2021年就已开展合作。2021年9月，商汤科技就与哪吒汽车达成了全面战略合作，共同在 智能驾驶和智能座舱领域展开深入合作。 目前，王俊平在内部架构中还 ...

SenseTime Group (00020) H1 2024 Earnings Call August 27, 2024 04:00 AM ET Speaker0 Hi, everyone. Welcome to Sensetime Group's twenty twenty four Mid Year Results Briefing. Our host is Sheng Shui. Joining us today from management are doctor Xu Li, chairman and CEO of Sun Time, mister Xu Bin, cofounder and executive director, Mr. Wang Ge, Chief Financial Officer. Let me begin by reading the disclaimer. This discussion may contain forward looking statements. Forward looking statements involve inherent risks an ...