NestPipe: Large-Scale Recommendation Training on 1,500+ Accelerators via Nested Pipelining

Modern recommendation models have increased to trillions of parameters. As cluster scales expand to O(1k), distributed training bottlenecks shift from computation and memory to data movement, especially lookup and communication latency associated with embeddings. Existing solutions either optimize only one bottleneck or improve throughput by sacrificing training consistency. This paper presents NestPipe, a large-scale decentralized embedding training framework that tackles both bottlenecks while preserving synchronous training semantics. NestPipe exploits two hierarchical sparse parallelism opportunities through nested pipelining. At the inter-batch level, Dual-Buffer Pipelining (DBP) constructs a staleness-free five-stage pipeline through dual-buffer synchronization, mitigating lookup bottlenecks without embedding staleness. At the intra-batch level, we identify the embedding freezing phenomenon, which inspires Frozen-Window Pipelining (FWP) to overlap All2All communication with dense computation via coordinated stream scheduling and key-centric sample clustering. Experiments on production GPU and NPU clusters with 1,536 workers demonstrate that NestPipe achieves up to 3.06x speedup and 94.07% scaling efficiency.

ClickAIXR: On-Device Multimodal Vision-Language Interaction with Real-World Objects in Extended Reality

We present ClickAIXR, a novel on-device framework for multimodal vision-language interaction with objects in extended reality (XR). Unlike prior systems that rely on cloud-based AI (e.g., ChatGPT) or gaze-based selection (e.g., GazePointAR), ClickAIXR integrates an on-device vision-language model (VLM) with a controller-based object selection paradigm, enabling users to precisely click on real-world objects in XR. Once selected, the object image is processed locally by the VLM to answer natural language questions through both text and speech. This object-centered interaction reduces ambiguity inherent in gaze- or voice-only interfaces and improves transparency by performing all inference on-device, addressing concerns around privacy and latency. We implemented ClickAIXR in the Magic Leap SDK (C API) with ONNX-based local VLM inference. We conducted a user study comparing ClickAIXR with Gemini 2.5 Flash and ChatGPT 5, evaluating usability, trust, and user satisfaction. Results show that latency is moderate and user experience is acceptable. Our findings demonstrate the potential of click-based object selection combined with on-device AI to advance trustworthy, privacy-preserving XR interactions. The source code and supplementary materials are available at: nanovis.org/ClickAIXR.html

Harnessing Lightweight Transformer with Contextual Synergic Enhancement for Efficient 3D Medical Image Segmentation

Transformers have shown remarkable performance in 3D medical image segmentation, but their high computational requirements and need for large amounts of labeled data limit their applicability. To address these challenges, we consider two crucial aspects: model efficiency and data efficiency. Specifically, we propose Light-UNETR, a lightweight transformer designed to achieve model efficiency. Light-UNETR features a Lightweight Dimension Reductive Attention (LIDR) module, which reduces spatial and channel dimensions while capturing both global and local features via multi-branch attention. Additionally, we introduce a Compact Gated Linear Unit (CGLU) to selectively control channel interaction with minimal parameters. Furthermore, we introduce a Contextual Synergic Enhancement (CSE) learning strategy, which aims to boost the data efficiency of Transformers. It first leverages the extrinsic contextual information to support the learning of unlabeled data with Attention-Guided Replacement, then applies Spatial Masking Consistency that utilizes intrinsic contextual information to enhance the spatial context reasoning for unlabeled data. Extensive experiments on various benchmarks demonstrate the superiority of our approach in both performance and efficiency. For example, with only 10% labeled data on the Left Atrial Segmentation dataset, our method surpasses BCP by 1.43% Jaccard while drastically reducing the FLOPs by 90.8% and parameters by 85.8%. Code is released at https://github.com/CUHK-AIM-Group/Light-UNETR.

Model Selection and Parameter Estimation of Multi-dimensional Gaussian Mixture Model

In this paper, we study the problem of learning multi-dimensional Gaussian Mixture Models (GMMs), with a specific focus on model order selection and efficient mixing distribution estimation. We first establish an information-theoretic lower bound on the critical sample complexity required for reliable model selection. More specifically, we show that distinguishing a $k$-component mixture from a simpler model necessitates a sample size scaling of $Ω(Δ^{-(4k-4)})$. We then propose a thresholding-based estimation algorithm that evaluates the spectral gap of an empirical covariance matrix constructed from random Fourier measurement vectors. This parameter-free estimator operates with an efficient time complexity of $\mathcal{O}(k^2 n)$, scaling linearly with the sample size. We demonstrate that the sample complexity of our method matches the established lower bound, confirming its minimax optimality with respect to the component separation distance $Δ$. Conditioned on the estimated model order, we subsequently introduce a gradient-based minimization method for parameter estimation. To effectively navigate the non-convex objective landscape, we employ a data-driven, score-based initialization strategy that guarantees rapid convergence. We prove that this method achieves the optimal parametric convergence rate of $\mathcal{O}_p(n^{-1/2})$ for estimating the component means. To enhance the algorithm's efficiency in high-dimensional regimes where the ambient dimension exceeds the number of mixture components (i.e., \(d > k\)), we integrate principal component analysis (PCA) for dimension reduction. Numerical experiments demonstrate that our Fourier-based algorithmic framework outperforms conventional Expectation-Maximization (EM) methods in both estimation accuracy and computational time.

DreamVideo-Omni: Omni-Motion Controlled Multi-Subject Video Customization with Latent Identity Reinforcement Learning

While large-scale diffusion models have revolutionized video synthesis, achieving precise control over both multi-subject identity and multi-granularity motion remains a significant challenge. Recent attempts to bridge this gap often suffer from limited motion granularity, control ambiguity, and identity degradation, leading to suboptimal performance on identity preservation and motion control. In this work, we present DreamVideo-Omni, a unified framework enabling harmonious multi-subject customization with omni-motion control via a progressive two-stage training paradigm. In the first stage, we integrate comprehensive control signals for joint training, encompassing subject appearances, global motion, local dynamics, and camera movements. To ensure robust and precise controllability, we introduce a condition-aware 3D rotary positional embedding to coordinate heterogeneous inputs and a hierarchical motion injection strategy to enhance global motion guidance. Furthermore, to resolve multi-subject ambiguity, we introduce group and role embeddings to explicitly anchor motion signals to specific identities, effectively disentangling complex scenes into independent controllable instances. In the second stage, to mitigate identity degradation, we design a latent identity reward feedback learning paradigm by training a latent identity reward model upon a pretrained video diffusion backbone. This provides motion-aware identity rewards in the latent space, prioritizing identity preservation aligned with human preferences. Supported by our curated large-scale dataset and the comprehensive DreamOmni Bench for multi-subject and omni-motion control evaluation, DreamVideo-Omni demonstrates superior performance in generating high-quality videos with precise controllability.

Rollout-Training Co-Design for Efficient LLM-Based Multi-Agent Reinforcement Learning

Despite algorithm-level innovations for multi-agent reinforcement learning (MARL), the underlying networked infrastructure for large-scale MARL training remains underexplored. Existing training frameworks primarily optimize for single-agent scenarios and fail to address the unique system-level challenges of MARL, including rollout-training synchronization barriers, rollout load imbalance, and training resource underutilization. To bridge this gap, we propose FlexMARL, the first end-to-end training framework that holistically optimizes rollout, training, and their orchestration for large-scale LLM-based MARL. Specifically, FlexMARL introduces the joint orchestrator to manage data flow under the rollout-training disaggregated architecture. Building upon the experience store, a novel micro-batch driven asynchronous pipeline eliminates the synchronization barriers while providing strong consistency guarantees. Rollout engine adopts a parallel sampling scheme combined with hierarchical load balancing, which adapts to skewed inter/intra-agent request patterns. Training engine achieves on-demand hardware binding through agent-centric resource allocation. The training states of different agents are swapped via unified and location-agnostic communication. Empirical results on a large-scale production cluster demonstrate that FlexMARL achieves up to 7.3x speedup and improves hardware utilization by up to 5.6x compared to existing frameworks.

DINO-Mix: Distilling Foundational Knowledge with Cross-Domain CutMix for Semi-supervised Class-imbalanced Medical Image Segmentation

Semi-supervised learning (SSL) has emerged as a critical paradigm for medical image segmentation, mitigating the immense cost of dense annotations. However, prevailing SSL frameworks are fundamentally "inward-looking", recycling information and biases solely from within the target dataset. This design triggers a vicious cycle of confirmation bias under class imbalance, leading to the catastrophic failure to recognize minority classes. To dismantle this systemic issue, we propose a paradigm shift to a multi-level "outward-looking" framework. Our primary innovation is Foundational Knowledge Distillation (FKD), which looks outward beyond the confines of medical imaging by introducing a pre-trained visual foundation model, DINOv3, as an unbiased external semantic teacher. Instead of trusting the student's biased high confidence, our method distills knowledge from DINOv3's robust understanding of high semantic uniqueness, providing a stable, cross-domain supervisory signal that anchors the learning of minority classes. To complement this core strategy, we further look outward within the data by proposing Progressive Imbalance-aware CutMix (PIC), which creates a dynamic curriculum that adaptively forces the model to focus on minority classes in both labeled and unlabeled subsets. This layered strategy forms our framework, DINO-Mix, which breaks the vicious cycle of bias and achieves remarkable performance on challenging semi-supervised class-imbalanced medical image segmentation benchmarks Synapse and AMOS.

Understanding Real-World Traffic Safety through RoadSafe365 Benchmark

Although recent traffic benchmarks have advanced multimodal data analysis, they generally lack systematic evaluation aligned with official safety standards. To fill this gap, we introduce RoadSafe365, a large-scale vision-language benchmark that supports fine-grained analysis of traffic safety from extensive and diverse real-world video data collections. Unlike prior works that focus primarily on coarse accident identification, RoadSafe365 is independently curated and systematically organized using a hierarchical taxonomy that refines and extends foundational definitions of crash, incident, and violation to bridge official traffic safety standards with data-driven traffic understanding systems. RoadSafe365 provides rich attribute annotations across diverse traffic event types, environmental contexts, and interaction scenarios, yielding 36,196 annotated clips from both dashcam and surveillance cameras. Each clip is paired with multiple-choice question-answer sets, comprising 864K candidate options, 8.4K unique answers, and 36K detailed scene descriptions collectively designed for vision-language understanding and reasoning. We establish strong baselines and observe consistent gains when fine-tuning on RoadSafe365. Cross-domain experiments on both real and synthetic datasets further validate its effectiveness. Designed for large-scale training and standardized evaluation, RoadSafe365 provides a comprehensive benchmark to advance reproducible research in real-world traffic safety analysis.

MathlibLemma: Folklore Lemma Generation and Benchmark for Formal Mathematics

While the ecosystem of Lean and Mathlib has enjoyed celebrated success in formal mathematical reasoning with the help of large language models (LLMs), the absence of many folklore lemmas in Mathlib remains a persistent barrier that limits Lean's usability as an everyday tool for mathematicians like LaTeX or Maple. To address this, we introduce MathlibLemma, the first LLM-based multi-agent system to automate the discovery and formalization of mathematical folklore lemmas. This framework constitutes our primary contribution, proactively mining the missing connective tissue of mathematics. Its efficacy is demonstrated by the production of a verified library of folklore lemmas, a subset of which has already been formally merged into the latest build of Mathlib, thereby validating the system's real-world utility and alignment with expert standards. Leveraging this pipeline, we further construct the MathlibLemma benchmark, a suite of 4,028 type-checked Lean statements spanning a broad range of mathematical domains. By transforming the role of LLMs from passive consumers to active contributors, this work establishes a constructive methodology for the self-evolution of formal mathematical libraries.

SpanNorm: Reconciling Training Stability and Performance in Deep Transformers

The success of Large Language Models (LLMs) hinges on the stable training of deep Transformer architectures. A critical design choice is the placement of normalization layers, leading to a fundamental trade-off: the ``PreNorm'' architecture ensures training stability at the cost of potential performance degradation in deep models, while the ``PostNorm'' architecture offers strong performance but suffers from severe training instability. In this work, we propose SpanNorm, a novel technique designed to resolve this dilemma by integrating the strengths of both paradigms. Structurally, SpanNorm establishes a clean residual connection that spans the entire transformer block to stabilize signal propagation, while employing a PostNorm-style computation that normalizes the aggregated output to enhance model performance. We provide a theoretical analysis demonstrating that SpanNorm, combined with a principled scaling strategy, maintains bounded signal variance throughout the network, preventing the gradient issues that plague PostNorm models, and also alleviating the representation collapse of PreNorm. Empirically, SpanNorm consistently outperforms standard normalization schemes in both dense and Mixture-of-Experts (MoE) scenarios, paving the way for more powerful and stable Transformer architectures.