你也被刷屏了吧？向来低调的阶跃星辰一口气抛出了两个深水炸弹。 一是完成超50亿元人民币的B+轮融资。 这个数字刷新⼤模型赛道过去12个月单笔融资纪录，且超过大模型六小虎中另外两家——智谱和MiniMax——的IPO募资金额。 另一个更具结构性变化的消息也在此时公布：印奇正式加入阶跃星辰核心决策层，担任董事长。 但对阶跃来说，豪华阵容只是最不值一提的表象。 能看懂这个班底背后的故事，就不难看懂阶跃为什么从以前，到现在，都反复强调并坚持自己"AI+终端"的战略。 印奇是谁？或许能用几个tag来标记他至今的传奇路径。 清华姚班首批校友，AI四小龙旷视科技创始人，现千里科技董事长。 是中国AI创业史上一位极少数横跨AI 1.0时代（CV）与AI 2.0（大模型）时代，同时又完成过产业落地与硬件闭环的样本人物。 从这一刻起，阶跃的核心决策层班底，明晰呈现出"1+3"模式。 1：指新宣布加入的印奇； 3：分别是阶跃星辰CEO姜大昕，首席科学家张祥⾬，CTO朱亦博。 每一个人都大名鼎鼎，单拎出来都是一部行业简史。 借这个时间，借这个新动态，刚好和大家一起来起底一下阶跃星辰"1+3"战队背后的事儿。 起底阶跃"1+3"核心团 ...

衡宇 发自 凹非寺 量子位 | 公众号 QbitAI 你也被刷屏了吧？向来低调的阶跃星辰一口气抛出了两个深水炸弹。 一是完成超50亿元人民币的B+轮融资 。 这个数字刷新⼤模型赛道过去12个月单笔融资纪录，且超过大模型六小虎中另外两家——智谱和MiniMax——的IPO募资金额。 另一个更具结构性变化的消息也在此时公布： 印奇正式加入阶跃星辰核心决策层，担任董事长 。 印奇是谁？或许能用几个tag来标记他至今的传奇路径。 清华姚班首批校友，AI四小龙旷视科技创始人，现千里科技董事长。 是中国AI创业史上 一位 极少数横跨AI 1.0时代 （CV） 与AI 2.0 （大模型） 时代，同时又完成过产业落地与硬件闭环的样本人物。 从这一刻起， 阶跃的核心决策层班底，明晰呈现出"1+3"模式 。 每一个人都大名鼎鼎，单拎出来都是一部行业简史。 但对阶跃来说，豪华阵容只是最不值一提的表象。 能看懂这个班底背后的故事，就不难看懂阶跃为什么从以前，到现在，都反复强调并坚持自己"AI+终端"的战略 。 借这个时间，借这个新动态，刚好和大家一起来起底一下阶跃星辰"1+3"战队背后的事儿。 起底阶跃"1+3"核心团队 1： 指新 ...

Group 1 - NeurIPS is recognized as the "Oscar of AI" and serves as a global annual barometer for the artificial intelligence field [1] - The NeurIPS Time-Tested Award honors foundational works that have significantly influenced the discipline over a decade [1] - The award was given to the authors of "Faster R-CNN," which has been cited over 98,000 times, making it the most cited paper by a Chinese first author at this conference [2] Group 2 - "Faster R-CNN," developed in 2015, improved object detection efficiency by over 10 times and introduced an end-to-end real-time detection model [2] - The core ideas of this model have been deeply integrated into the foundational technologies of AI, impacting key sectors such as autonomous driving and medical imaging [2] - The collaboration between the authors, including Ren Shaoqing and He Kaiming, has led to significant advancements in deep learning frameworks [2] Group 3 - Ren Shaoqing joined NIO in August 2020, focusing on building a team and developing self-research chips for autonomous driving [13][14] - NIO's first generation of vehicles utilized the Mobileye solution, while the second generation was the first globally to mass-produce the NVIDIA Orin chip [14] - The challenges faced during the development included adapting to new architectures and ensuring the stability of the new chip [15] Group 4 - NIO emphasized the importance of data collection and analysis, focusing on corner cases to improve the performance of their models [19][20] - The company established a flexible system for cloud computing and data management, allowing for rapid iteration of models [21] - NIO's approach to active safety has enabled them to achieve a standard of 200,000 kilometers per false positive, significantly improving their testing efficiency [22] Group 5 - The concept of end-to-end solutions in autonomous driving has evolved, with discussions on integrating various technologies to enhance performance [24][25] - NIO is exploring the development of world models to improve long-term decision-making capabilities in autonomous systems [27][28] - The world model approach aims to address the limitations of traditional methods by incorporating both spatial and temporal understanding [30][31]

Core Insights - The article highlights the recruitment of two new Chinese students, Hu Keya and Li Zongyi, by AI expert He Kaiming at MIT, emphasizing their impressive academic backgrounds and contributions to the field of AI [1][4]. Group 1: Hu Keya's Background and Achievements - Hu Keya graduated from Shanghai Jiao Tong University and was involved in the Brain-Machine Interface Laboratory, focusing on AI applications in neuroscience [5][7]. - She authored a paper on self-supervised EEG representation learning, which was accepted at the EMBC conference, and presented her work in the U.S. [8][10]. - Hu participated in a project that improved self-supervised learning, leading to a paper accepted at the Cognitive Science 2025 conference [10]. - During her undergraduate studies, she interned at Cornell University, contributing to a project on program synthesis and code repair, resulting in a paper accepted at NeurIPS 2024 [11][12]. - Hu Keya led her team to win the "Best Paper Award" at the ARC Prize 2024 competition, showcasing her innovative approach to AI problem-solving [15][17]. - By the end of her undergraduate studies, she had published four high-impact papers, making her a highly sought-after candidate for PhD programs, ultimately choosing MIT [21][22]. Group 2: Li Zongyi's Contributions - Li Zongyi, known for his work on the Fourier Neural Operator (FNO), published a significant paper during his PhD that enabled the large-scale application of neural operators [27][29]. - The FNO allows neural networks to learn solutions to physical equations efficiently, significantly improving computational speed in various scientific applications [30][34]. - Li Zongyi's research has made him a key figure in the field of neural operators, with over 12,000 citations of his work [36]. - Currently, he is a postdoctoral researcher at MIT and is set to join New York University as an assistant professor in the upcoming fall [38][39]. Group 3: He Kaiming's Research Focus - He Kaiming has indicated that "AI for Science" will be a primary focus of his research in the coming years, aligning with the expertise of his newly recruited team members [46][48]. - The combination of Hu Keya's background in neuroscience and Li Zongyi's expertise in neural operators strengthens the team's capabilities in advancing AI applications in scientific research [48][49].

Model Calibration Importance - Modern neural networks can be misleading due to overconfidence in predictions [1][2] - Calibration ensures predicted probabilities align with actual outcomes, crucial for reliable decision-making [2][3] - Overly confident but inaccurate models can lead to suboptimal decisions, exemplified by unnecessary medical tests [3] Calibration Assessment - Reliability Diagrams visually inspect model calibration by plotting expected accuracy against confidence [4] - Expected Calibration Error (ECE) quantifies miscalibration, approximated by averaging accuracy/confidence differences across bins [6] Calibration Techniques - Calibration is important when probabilities matter and models are operationally similar [7] - Binary classification models can be calibrated using histogram binning, isotonic regression, or Platt scaling [7] - Multiclass classification models can be calibrated using binning methods or matrix and vector scaling [7] Experimental Results - LeNet model achieved an accuracy of approximately 55% with an average confidence of approximately 54% [5] - ResNet model achieved an accuracy of approximately 70% but with a higher average confidence of approximately 90%, indicating overconfidence [5] - ResNet model thinks it's 90% confident in its predictions, in reality, it only turns out to be 70% accurate [2]

Core Viewpoint - The article discusses the historical context and contributions of Sepp Hochreiter and Jürgen Schmidhuber in the development of residual learning and its impact on deep learning, emphasizing that the concept of residual connections was introduced by Hochreiter in 1991, long before its popularization in ResNet [3][12][26]. Group 1: Historical Contributions - Sepp Hochreiter systematically analyzed the vanishing gradient problem in his 1991 doctoral thesis and proposed the use of recurrent residual connections to address this issue [3][12]. - The core idea of recurrent residual connections involves a self-connecting neuron with a fixed weight of 1.0, allowing the error signal to remain constant during backpropagation [13][14]. - The introduction of LSTM in 1997 by Hochreiter and Schmidhuber built upon this foundational concept, enabling effective long-term dependency learning in tasks such as speech and language processing [18][19]. Group 2: Evolution of Residual Learning - The Highway network, introduced in 2015, successfully trained deep feedforward networks with hundreds of layers by incorporating the gated residual concept from LSTM [23]. - ResNet, which gained significant attention in the same year, utilized residual connections to stabilize error propagation in deep networks, allowing for the training of networks with hundreds of layers [24][26]. - Both Highway networks and ResNet share similarities with the foundational principles established by Hochreiter in 1991, demonstrating the enduring relevance of his contributions to deep learning [26]. Group 3: Ongoing Debates and Recognition - Jürgen Schmidhuber has publicly claimed that various architectures, including AlexNet, VGG Net, GANs, and Transformers, were inspired by his lab's work, although these claims have not been universally accepted [28][31]. - The ongoing debate regarding the attribution of contributions in deep learning highlights the complexities of recognizing foundational work in a rapidly evolving field [10][32].

Core Viewpoint - Ren Shaoqing, a prominent figure in AI and autonomous driving, has returned to his alma mater, the University of Science and Technology of China, to start a new academic program focusing on advanced AI topics [4][6]. Group 1: Background of Ren Shaoqing - Ren Shaoqing is a co-founder of Momenta and former Vice President of NIO, with a strong academic background including a PhD from the University of Science and Technology of China [4]. - He is recognized for his contributions to AI, particularly as the author of ResNet and Faster R-CNN, with over 440,000 citations, making him the most cited Chinese scholar globally [4]. Group 2: Academic Program Details - The new program will focus on areas such as AGI (Artificial General Intelligence), world models, embodied intelligence, and AI for Science [6]. - The program is open for recruitment of master's and doctoral students, with urgent interviews scheduled for students with recommendation qualifications starting next Monday [6].

Core Viewpoint - Ren Shaoqing, a prominent figure in AI and computer vision, is starting a recruitment program at his alma mater, the University of Science and Technology of China, focusing on advanced topics in AI such as AGI, world models, embodied intelligence, and AI for Science [1][2]. Group 1: Recruitment Details - The recruitment is open for both master's and doctoral students, with emergency interviews starting on the upcoming Monday for students with recommendation qualifications [3]. - Interested students can send their resumes to Ren Shaoqing's email for inquiries regarding the application process and interview details [16]. Group 2: Background of Ren Shaoqing - Ren Shaoqing is an expert in computer vision and autonomous driving, having graduated from the University of Science and Technology of China and obtained a joint PhD with Microsoft Research Asia [4][5]. - He has been recognized as one of the most influential scholars in AI, ranking 10th in the AI 2000 list, and received the Future Science Prize in Mathematics and Computer Science in 2023 [6]. Group 3: Contributions to AI - Ren is a co-author of ResNet, a groundbreaking work in deep learning that addresses the vanishing gradient problem, significantly impacting fields requiring high perception capabilities like computer vision and autonomous driving [7]. - ResNet has received over 290,000 citations and won the Best Paper Award at CVPR 2016 [8]. - He also contributed to Faster R-CNN, an efficient two-stage object detection algorithm that balances speed and accuracy [10]. Group 4: Role in NIO - After completing his PhD, Ren co-founded Momenta and later joined NIO, where he played a key role in developing autonomous driving algorithms and leading the smart driving R&D team [13]. - At NIO, he developed the NIO World Model (NWM), which integrates spatiotemporal cognition and generative capabilities, allowing for high-fidelity scene reconstruction and long-term scenario simulation [14][15].

Core Insights - Yoshua Bengio has become the most cited scientist in history with a total citation count of 973,655 and 698,008 citations in the last five years [1] - The ranking is based on total citation counts and recent citation indices from AD Scientific Index, which evaluates scientists across various disciplines [1] - Bengio's work on Generative Adversarial Networks (GANs) has surpassed 100,000 citations, indicating significant impact in the AI field [1] Group 1 - The second-ranked scientist is Geoffrey Hinton, with over 950,000 total citations and more than 570,000 citations in the last five years [3] - Hinton's collaboration on the AlexNet paper has received over 180,000 citations, marking a pivotal moment in deep learning for computer vision [3] - The third and fourth positions in the citation rankings are held by researchers in the medical field, highlighting the interdisciplinary nature of high-impact research [6] Group 2 - Kaiming He ranks fifth, with his paper on Deep Residual Learning for Image Recognition cited over 290,000 times, establishing a foundation for modern deep learning [6] - The paper by He is recognized as the most cited paper of the 21st century according to Nature, emphasizing its lasting influence [9] - Ilya Sutskever, another prominent figure in AI, ranks seventh with over 670,000 total citations, showcasing the strong presence of AI researchers in citation rankings [10]

Core Insights - The article highlights the prominence of AI as the hottest research direction globally, with Yoshua Bengio being the most cited scientist ever, accumulating a total citation count of 973,655 and 698,008 citations in the last five years [1][3]. Group 1: Citation Rankings - The AD Scientific Index ranks 2,626,749 scientists from 221 countries and 24,576 institutions based on total citation counts and recent citation indices [3]. - Yoshua Bengio's work on Generative Adversarial Networks (GANs) has surpassed 100,000 citations, outpacing his co-authored paper "Deep Learning," which also exceeds 100,000 citations [3][4]. - Geoffrey Hinton, a pioneer in AI, ranks second with over 950,000 total citations and more than 570,000 citations in the last five years [4][5]. Group 2: Notable Papers and Their Impact - The paper "AlexNet," co-authored by Hinton, Krizhevsky, and Sutskever, has received over 180,000 citations, marking a significant breakthrough in deep learning for computer vision [5][6]. - Kaiming He’s paper "Deep Residual Learning for Image Recognition" has over 290,000 citations, establishing ResNet as a foundational model in modern deep learning [10][11]. - The article notes that ResNet is recognized as the most cited paper of the 21st century, with citation counts ranging from 103,756 to 254,074 across various databases [11]. Group 3: Broader Implications - The high citation counts of these influential papers indicate their lasting impact on the academic community and their role in shaping future research directions in AI and related fields [17].