Jialin Li

Xuechao Chen, Ying Chen, Jialin Li et al.

3D pre-training is crucial to 3D perception tasks. Nevertheless, limited by the difficulties in collecting clean and complete 3D data, 3D pre-training has persistently faced data scaling challenges. In this work, we introduce a novel self-supervised pre-training framework that incorporates millions of images into 3D pre-training corpora by leveraging a large depth estimation model. New pre-training corpora encounter new challenges in representation ability and embedding efficiency of models. Previous pre-training methods rely on farthest point sampling and k-nearest neighbors to embed a fixed number of 3D tokens. However, these approaches prove inadequate when it comes to embedding millions of samples that feature a diverse range of point numbers, spanning from 1,000 to 100,000. In contrast, we propose a tokenizer with linear-time complexity, which enables the efficient embedding of a flexible number of tokens. Accordingly, a new 3D reconstruction target is proposed to cooperate with our 3D tokenizer. Our method achieves state-of-the-art performance in 3D classification, few-shot learning, and 3D segmentation. Code is available at https://github.com/XuechaoChen/P3P-MAE.

Zhouhong Gu, Xiaoxuan Zhu, Yin Cai et al.

Large language model based multi-agent systems have demonstrated significant potential in social simulation and complex task resolution domains. However, current frameworks face critical challenges in system architecture design, cross-domain generalizability, and performance guarantees, particularly as task complexity and number of agents increases. We introduces AgentGroupChat-V2, a novel framework addressing these challenges through three core innovations: (1) a divide-and-conquer fully parallel architecture that decomposes user queries into hierarchical task forest structures enabling dependency management and distributed concurrent processing. (2) an adaptive collaboration engine that dynamically selects heterogeneous LLM combinations and interaction modes based on task characteristics. (3) agent organization optimization strategies combining divide-and-conquer approaches for efficient problem decomposition. Extensive experiments demonstrate AgentGroupChat-V2's superior performance across diverse domains, achieving 91.50% accuracy on GSM8K (exceeding the best baseline by 5.6 percentage points), 30.4% accuracy on competition-level AIME (nearly doubling other methods), and 79.20% pass@1 on HumanEval. Performance advantages become increasingly pronounced with higher task difficulty, particularly on Level 5 MATH problems where improvements exceed 11 percentage points compared to state-of-the-art baselines. These results confirm that AgentGroupChat-V2 provides a comprehensive solution for building efficient, general-purpose LLM multi-agent systems with significant advantages in complex reasoning scenarios. Code is available at https://github.com/MikeGu721/AgentGroupChat-V2.

Xinyi Wan, Penghui Qi, Guangxing Huang et al.

Pipeline parallelism (PP) is widely used for training large language models (LLMs), yet its scalability is often constrained by high activation memory consumption as the number of in-flight microbatches grows with the degree of PP. In this paper, we focus on addressing this challenge by leveraging the under-explored memory offload strategy in PP. With empirical study, we discover that in the majority of standard configurations, at least half, and potentially all, of the activations can be offloaded with negligible overhead. In the cases where full overload is not possible, we introduce a novel selective offload strategy that decreases peak activation memory in a better-than-linear manner. Furthermore, we integrate memory offload with other techniques to jointly consider overall throughput and memory limitation. Our experiments proves that the per-device activation memory effectively reduces with the total number of stages, making PP a stronger alternative than TP, offering up to a 19\% acceleration with even lower memory consumption. The implementation is open-sourced at \href{https://github.com/sail-sg/zero-bubble-pipeline-parallelism}{this url}.

R. Landman, S. Y. Haffert, J. D. Long et al.

Pyramid wavefront sensors (PWFSs) are the preferred choice for current and future extreme adaptive optics (XAO) systems. Almost all instruments use the PWFS in its modulated form to mitigate its limited linearity range. However, this modulation comes at the cost of a reduction in sensitivity, a blindness to petal-piston modes, and a limit to the sensor's ability to operate at high speeds. Therefore, there is strong interest to use the PWFS without modulation, which can be enabled with nonlinear reconstructors. Here, we present the first on-sky demonstration of XAO with an unmodulated PWFS using a nonlinear reconstructor based on convolutional neural networks. We discuss the real-time implementation on the Magellan Adaptive Optics eXtreme (MagAO-X) instrument using the optimized TensorRT framework and show that inference is fast enough to run the control loop at >2 kHz frequencies. Our on-sky results demonstrate a successful closed-loop operation using a model calibrated with internal source data that delivers stable and robust correction under varying conditions. Performance analysis reveals that our smart PWFS achieves nearly the same Strehl ratio as the highly optimized modulated PWFS under favorable conditions on bright stars. Notably, we observe an improvement in performance on a fainter star under the influence of strong winds. These findings confirm the feasibility of using the PWFS in its unmodulated form and highlight its potential for next-generation instruments. Future efforts will focus on achieving even higher control loop frequencies (>3 kHz), optimizing the calibration procedures, and testing its performance on fainter stars, where more gain is expected for the unmodulated PWFS compared to its modulated counterpart.