Core Viewpoint - The article discusses the recent leadership changes in XPeng Motors' autonomous driving division, highlighting the appointment of Xianming Liu as the new head of the autonomous driving center, and the strategic focus on AI large models for enhancing autonomous driving capabilities [5][7][9]. Group 1: Leadership Changes - On October 9, XPeng Motors announced that Li Liyun would no longer lead the autonomous driving center, with Xianming Liu taking over the role [5]. - Xianming Liu has a strong background in machine learning and computer vision, having previously worked at Facebook and Cruise, and joined XPeng in March 2024 [5][7]. - Li Liyun, who took over the autonomous driving team in August 2023, has a background in electronic engineering and computer science, and has been instrumental in the development of XPeng's intelligent driving solutions [8][9]. Group 2: Strategic Focus on AI - The autonomous driving VLA model is a significant application of the world base model, which aims to enhance the predictive capabilities of autonomous driving systems [7]. - XPeng has been developing a large-scale autonomous driving model with 72 billion parameters, marking a significant investment in AI technology [9][10]. - The company plans to invest 4.5 billion yuan in AI and autonomous driving by 2025, indicating a commitment to advancing its technological capabilities [10]. Group 3: Industry Context - The competition in the intelligent driving sector among Chinese automakers has intensified, with companies like Li Auto and NIO making significant advancements in their autonomous driving technologies [9][10]. - The article notes that 2023 has seen significant personnel changes in the intelligent driving teams across various Chinese automotive companies, reflecting a broader trend in the industry [10][11].

Core Viewpoint - The article discusses the recent leadership changes in Xiaopeng Motors' autonomous driving team, highlighting the appointment of Xianming Liu as the new head of the autonomous driving center, and the strategic focus on AI large models for enhancing autonomous driving capabilities [3][4][6]. Group 1: Leadership Changes - Xianming Liu has replaced Li Liyun as the head of Xiaopeng's autonomous driving center, with a strong background in machine learning and computer vision from previous roles at Facebook and Cruise [3][4]. - Li Liyun, who took over the autonomous driving team in August 2023, has a notable background in technology and has been instrumental in the development of Xiaopeng's intelligent driving solutions [6][7]. Group 2: Strategic Focus on AI - Xiaopeng is focusing on the development of a "world base model" for autonomous driving, which is a significant application of their AI large model initiative [6][8]. - The company has announced plans to develop a super-large autonomous driving model with 72 billion parameters, aiming to enhance its competitive edge in the intelligent driving sector [8][9]. Group 3: Industry Context - The competitive landscape in the intelligent driving sector is intensifying, with other major players like Li Auto, Huawei, and NIO also ramping up their AI and autonomous driving technology efforts [8][9]. - Xiaopeng has established the first large-scale AI computing cluster in the domestic automotive industry, supporting the training of its various models [9][10].

Core Viewpoint - The article discusses the breakthrough research VLA-OS by a team from the National University of Singapore, which systematically analyzes and dissects the task planning and reasoning of Vision-Language-Action (VLA) models, providing a clear direction for the next generation of general-purpose robotic VLA models [3][5]. Group 1: VLA Model Analysis - VLA models have shown impressive capabilities in solving complex tasks through end-to-end data-driven imitation learning, mapping raw image and language inputs directly to robotic action spaces [9][11]. - Current datasets for training VLA models are limited compared to those for Large Language Models (LLMs) and Vision-Language Models (VLMs), prompting researchers to integrate task reasoning modules to enhance performance with less data [11][12]. - The article identifies two main approaches for integrating task reasoning: Integrated-VLA, which combines task planning and strategy learning, and Hierarchical-VLA, which separates these functions into different models [12][13]. Group 2: VLA-OS Framework - VLA-OS serves as a modular experimental platform for VLA models, allowing for controlled variable experiments focused on task planning paradigms and representations [22][23]. - The framework includes a unified architecture with a family of VLM models, designed to facilitate fair comparisons among different VLA paradigms [23][25]. - A comprehensive multimodal task planning dataset has been created, covering various dimensions such as visual modalities, operational environments, and types of manipulators, totaling approximately 10,000 trajectories [28][29]. Group 3: Findings and Insights - The research yielded 14 valuable findings, highlighting the advantages of visual planning representations over language-based ones and the potential of hierarchical VLA paradigms for future development [35][36]. - Performance tests on the VLA-OS model showed that it outperformed several existing VLA models, indicating its competitive design even without pre-training [37][38]. - The study found that implicit task planning in Integrated-VLA models outperformed explicit planning, suggesting that auxiliary task planning objectives can enhance model performance [40][44]. Group 4: Recommendations and Future Directions - The article provides design guidelines, recommending the use of visual planning and goal image planning as primary methods, with language planning as a supplementary approach [81][82]. - It emphasizes the importance of task planning pre-training and suggests that hierarchical VLA models should be prioritized when resources allow [83][84]. - Future research directions include exploring the neural mechanisms behind spatial representations, developing more efficient VLM information distillation architectures, and constructing large-scale planning datasets for robotic operations [86].

Investment Rating - The report maintains a "Buy" rating for the company [5][7]. Core Views - The company is expected to achieve a gross margin of over 20.5% in Q1 2025, exceeding expectations despite a significant decrease in sales volume [1]. - The company anticipates delivering between 123,000 to 128,000 vehicles in Q2, representing a year-on-year growth of 13.3% to 17.9% [2]. - The launch of the first pure electric SUV, i8, is scheduled for July, with the VLA model also expected to be released [2][4]. Financial Performance - In Q1, the company sold 93,000 vehicles, with revenue reaching 25.93 billion RMB, reflecting a year-on-year increase of 1% [1]. - The Q1 net profit attributable to shareholders was 650 million RMB, with a net profit margin of 2.5% [1]. - The company projects total revenues of 155.5 billion RMB, 197.9 billion RMB, and 238.3 billion RMB for the years 2025, 2026, and 2027 respectively [5][6]. Future Outlook - The company plans to introduce more affordable MPV and sedan models based on market demand after the launch of the L series and i series [2]. - The MEGA Home model has been well-received, with over 90% of MEGA orders being for this version, indicating strong market insight and product definition capabilities [3]. - The company is expanding its supercharging network, with 2,328 supercharging stations and 12,689 supercharging piles nationwide, enhancing the appeal of its electric vehicles [4]. Sales and Production Forecast - The company expects to sell approximately 580,000, 750,000, and 870,000 vehicles in 2025, 2026, and 2027 respectively [5][6]. - The gross margin is projected to be around 19% in Q2 due to increased promotional efforts [2]. Valuation - The target market capitalization is set at 280.9 billion RMB, with a target price of approximately 131 HKD per share, corresponding to a 25x P/E ratio for 2025 [5].