arXiv subjects

Jiawei Yang

Publications and source records attributed to Jiawei Yang.

At least 37 records · Page 2Linked to original sources

arXiv2026

MiniCPM-SALA: Hybridizing Sparse and Linear Attention for Efficient Long-Context Modeling

The evolution of large language models (LLMs) towards applications with ultra-long contexts faces challenges posed by the high computational and memory costs of the Transformer architecture. While existing sparse and linear attention mechanisms attempt to mitigate these issues, they typically involve a trade-off between memory efficiency and model performance. This paper introduces MiniCPM-SALA, a 9B-parameter hybrid architecture that integrates the high-fidelity long-context modeling of sparse attention (InfLLM-V2) with the global efficiency of linear attention (Lightning Attention). By employing a layer selection algorithm to integrate these mechanisms in a 1:3 ratio and utilizing a hybrid positional encoding (HyPE), the model maintains efficiency and performance for long-context tasks. Furthermore, we introduce a cost-effective continual training framework that transforms pre-trained Transformer-based models into hybrid models, which reduces training costs by approximately 75% compared to training from scratch. Extensive experiments show that MiniCPM-SALA maintains general capabilities comparable to full-attention models while offering improved efficiency. On a single NVIDIA A6000D GPU, the model achieves up to 3.5x the inference speed of the full-attention model at the sequence length of 256K tokens and supports context lengths of up to 1M tokens, a scale where traditional full-attention 8B models fail because of memory constraints.

cs.CL↗

arXiv2026

Latent Denoising Makes Good Tokenizers

Despite their fundamental role, it remains unclear what properties could make tokenizers more effective for generative modeling. We observe that modern generative models share a conceptually similar training objective -- reconstructing clean signals from corrupted inputs, such as signals degraded by Gaussian noise or masking -- a process we term denoising. Motivated by this insight, we propose aligning tokenizer embeddings directly with the downstream denoising objective, encouraging latent embeddings that remain reconstructable even under significant corruption. To achieve this, we introduce the Latent Denoising Tokenizer (l-DeTok), a simple yet highly effective tokenizer trained to reconstruct clean images from latent embeddings corrupted via interpolative noise or random masking. Extensive experiments on class-conditioned (ImageNet 256x256 and 512x512) and text-conditioned (MSCOCO) image generation benchmarks demonstrate that our l-DeTok consistently improves generation quality across six representative generative models compared to prior tokenizers. Our findings highlight denoising as a fundamental design principle for tokenizer development, and we hope it could motivate new perspectives for future tokenizer design.

cs.CV↗

arXiv2025

Towards Efficient and Effective Multi-Camera Encoding for End-to-End Driving

We present Flex, an efficient and effective scene encoder that addresses the computational bottleneck of processing high-volume multi-camera data in end-to-end autonomous driving. Flex employs a small set of learnable scene tokens to jointly encode information from all image tokens across different cameras and timesteps. By design, our approach is geometry-agnostic, learning a compact scene representation directly from data without relying on the explicit 3D inductive biases, such as Bird-Eye-View (BEV), occupancy or tri-plane representations, which are common in prior work. This holistic encoding strategy aggressively compresses the visual input for the downstream Large Language Model (LLM) based policy model. Evaluated on a large-scale proprietary dataset of 20,000 driving hours, our Flex achieves 2.2x greater inference throughput while improving driving performance by a large margin compared to state-of-the-art methods. Furthermore, we show that these compact scene tokens develop an emergent capability for scene decomposition without any explicit supervision. Our findings challenge the prevailing assumption that 3D priors are necessary, demonstrating that a data-driven, joint encoding strategy offers a more scalable, efficient and effective path for future autonomous driving systems.

cs.CV↗

arXiv2025

MemOS: A Memory OS for AI System

Large Language Models (LLMs) have become an essential infrastructure for Artificial General Intelligence (AGI), yet their lack of well-defined memory management systems hinders the development of long-context reasoning, continual personalization, and knowledge consistency.Existing models mainly rely on static parameters and short-lived contextual states, limiting their ability to track user preferences or update knowledge over extended periods.While Retrieval-Augmented Generation (RAG) introduces external knowledge in plain text, it remains a stateless workaround without lifecycle control or integration with persistent representations.Recent work has modeled the training and inference cost of LLMs from a memory hierarchy perspective, showing that introducing an explicit memory layer between parameter memory and external retrieval can substantially reduce these costs by externalizing specific knowledge. Beyond computational efficiency, LLMs face broader challenges arising from how information is distributed over time and context, requiring systems capable of managing heterogeneous knowledge spanning different temporal scales and sources. To address this challenge, we propose MemOS, a memory operating system that treats memory as a manageable system resource. It unifies the representation, scheduling, and evolution of plaintext, activation-based, and parameter-level memories, enabling cost-efficient storage and retrieval. As the basic unit, a MemCube encapsulates both memory content and metadata such as provenance and versioning. MemCubes can be composed, migrated, and fused over time, enabling flexible transitions between memory types and bridging retrieval with parameter-based learning. MemOS establishes a memory-centric system framework that brings controllability, plasticity, and evolvability to LLMs, laying the foundation for continual learning and personalized modeling.

cs.CL↗

arXiv2025

Introducing LongCat-Flash-Thinking: A Technical Report

We present LongCat-Flash-Thinking, an efficient 560-billion-parameter open-source Mixture-of-Experts (MoE) reasoning model. Its advanced capabilities are cultivated through a meticulously crafted training process, beginning with long Chain-of-Thought (CoT) data cold-start and culminating in large-scale Reinforcement Learning (RL). We first employ a well-designed cold-start training strategy, which significantly enhances the reasoning potential and equips the model with specialized skills in both formal and agentic reasoning. Then, a core innovation is our domain-parallel training scheme, which decouples optimization across distinct domains (e.g., STEM, Code, Agentic) and subsequently fuses the resulting expert models into a single, nearly Pareto-optimal model. This entire process is powered by our Dynamic ORchestration for Asynchronous rollout (DORA) system, a large-scale RL framework that delivers a greater than threefold training speedup over synchronous methods on tens of thousands of accelerators. As a result, LongCat-Flash-Thinking achieves state-of-the-art performance among open-source models on a suite of complex reasoning tasks. The model exhibits exceptional efficiency in agentic reasoning, reducing average token consumption by 64.5% (from 19, 653 to 6, 965) on AIME-25, without degrading task accuracy. We release LongCat-Flash-Thinking to promote further advances in reasoning systems and agentic AI research.

cs.AI↗

arXiv2025

LongCat-Flash Technical Report

We introduce LongCat-Flash, a 560-billion-parameter Mixture-of-Experts (MoE) language model designed for both computational efficiency and advanced agentic capabilities. Stemming from the need for scalable efficiency, LongCat-Flash adopts two novel designs: (a) Zero-computation Experts, which enables dynamic computational budget allocation and activates 18.6B-31.3B (27B on average) per token depending on contextual demands, optimizing resource usage. (b) Shortcut-connected MoE, which enlarges the computation-communication overlap window, demonstrating notable gains in inference efficiency and throughput compared to models of a comparable scale. We develop a comprehensive scaling framework for large models that combines hyperparameter transfer, model-growth initialization, a multi-pronged stability suite, and deterministic computation to achieve stable and reproducible training. Notably, leveraging the synergy among scalable architectural design and infrastructure efforts, we complete model training on more than 20 trillion tokens within 30 days, while achieving over 100 tokens per second (TPS) for inference at a cost of \$0.70 per million output tokens. To cultivate LongCat-Flash towards agentic intelligence, we conduct a large-scale pre-training on optimized mixtures, followed by targeted mid- and post-training on reasoning, code, and instructions, with further augmentation from synthetic data and tool use tasks. Comprehensive evaluations demonstrate that, as a non-thinking foundation model, LongCat-Flash delivers highly competitive performance among other leading models, with exceptional strengths in agentic tasks. The model checkpoint of LongCat-Flash is open-sourced to foster community research. LongCat Chat: https://longcat.ai Hugging Face: https://huggingface.co/meituan-longcat GitHub: https://github.com/meituan-longcat

cs.CL↗

arXiv2025

BEVCon: Advancing Bird's Eye View Perception with Contrastive Learning

We present BEVCon, a simple yet effective contrastive learning framework designed to improve Bird's Eye View (BEV) perception in autonomous driving. BEV perception offers a top-down-view representation of the surrounding environment, making it crucial for 3D object detection, segmentation, and trajectory prediction tasks. While prior work has primarily focused on enhancing BEV encoders and task-specific heads, we address the underexplored potential of representation learning in BEV models. BEVCon introduces two contrastive learning modules: an instance feature contrast module for refining BEV features and a perspective view contrast module that enhances the image backbone. The dense contrastive learning designed on top of detection losses leads to improved feature representations across both the BEV encoder and the backbone. Extensive experiments on the nuScenes dataset demonstrate that BEVCon achieves consistent performance gains, achieving up to +2.4% mAP improvement over state-of-the-art baselines. Our results highlight the critical role of representation learning in BEV perception and offer a complementary avenue to conventional task-specific optimizations.

cs.CV↗

arXiv2025

Occupancy Learning with Spatiotemporal Memory

3D occupancy becomes a promising perception representation for autonomous driving to model the surrounding environment at a fine-grained scale. However, it remains challenging to efficiently aggregate 3D occupancy over time across multiple input frames due to the high processing cost and the uncertainty and dynamics of voxels. To address this issue, we propose ST-Occ, a scene-level occupancy representation learning framework that effectively learns the spatiotemporal feature with temporal consistency. ST-Occ consists of two core designs: a spatiotemporal memory that captures comprehensive historical information and stores it efficiently through a scene-level representation and a memory attention that conditions the current occupancy representation on the spatiotemporal memory with a model of uncertainty and dynamic awareness. Our method significantly enhances the spatiotemporal representation learned for 3D occupancy prediction tasks by exploiting the temporal dependency between multi-frame inputs. Experiments show that our approach outperforms the state-of-the-art methods by a margin of 3 mIoU and reduces the temporal inconsistency by 29%.

cs.CV↗

arXiv2025

InfiniCube: Unbounded and Controllable Dynamic 3D Driving Scene Generation with World-Guided Video Models

We present InfiniCube, a scalable method for generating unbounded dynamic 3D driving scenes with high fidelity and controllability. Previous methods for scene generation either suffer from limited scales or lack geometric and appearance consistency along generated sequences. In contrast, we leverage the recent advancements in scalable 3D representation and video models to achieve large dynamic scene generation that allows flexible controls through HD maps, vehicle bounding boxes, and text descriptions. First, we construct a map-conditioned sparse-voxel-based 3D generative model to unleash its power for unbounded voxel world generation. Then, we re-purpose a video model and ground it on the voxel world through a set of carefully designed pixel-aligned guidance buffers, synthesizing a consistent appearance. Finally, we propose a fast feed-forward approach that employs both voxel and pixel branches to lift the dynamic videos to dynamic 3D Gaussians with controllable objects. Our method can generate controllable and realistic 3D driving scenes, and extensive experiments validate the effectiveness and superiority of our model.

cs.CV↗

arXiv2025

MemOS: An Operating System for Memory-Augmented Generation (MAG) in Large Language Models

Large Language Models (LLMs) have emerged as foundational infrastructure in the pursuit of Artificial General Intelligence (AGI). Despite their remarkable capabilities in language perception and generation, current LLMs fundamentally lack a unified and structured architecture for handling memory. They primarily rely on parametric memory (knowledge encoded in model weights) and ephemeral activation memory (context-limited runtime states). While emerging methods like Retrieval-Augmented Generation (RAG) incorporate plaintext memory, they lack lifecycle management and multi-modal integration, limiting their capacity for long-term knowledge evolution. To address this, we introduce MemOS, a memory operating system designed for LLMs that, for the first time, elevates memory to a first-class operational resource. It builds unified mechanisms for representation, organization, and governance across three core memory types: parametric, activation, and plaintext. At its core is the MemCube, a standardized memory abstraction that enables tracking, fusion, and migration of heterogeneous memory, while offering structured, traceable access across tasks and contexts. MemOS establishes a memory-centric execution framework with strong controllability, adaptability, and evolvability. It fills a critical gap in current LLM infrastructure and lays the groundwork for continual adaptation, personalized intelligence, and cross-platform coordination in next-generation intelligent systems.

cs.CL↗

arXiv2025

OmniRe: Omni Urban Scene Reconstruction

We introduce OmniRe, a comprehensive system for efficiently creating high-fidelity digital twins of dynamic real-world scenes from on-device logs. Recent methods using neural fields or Gaussian Splatting primarily focus on vehicles, hindering a holistic framework for all dynamic foregrounds demanded by downstream applications, e.g., the simulation of human behavior. OmniRe extends beyond vehicle modeling to enable accurate, full-length reconstruction of diverse dynamic objects in urban scenes. Our approach builds scene graphs on 3DGS and constructs multiple Gaussian representations in canonical spaces that model various dynamic actors, including vehicles, pedestrians, cyclists, and others. OmniRe allows holistically reconstructing any dynamic object in the scene, enabling advanced simulations (~60Hz) that include human-participated scenarios, such as pedestrian behavior simulation and human-vehicle interaction. This comprehensive simulation capability is unmatched by existing methods. Extensive evaluations on the Waymo dataset show that our approach outperforms prior state-of-the-art methods quantitatively and qualitatively by a large margin. We further extend our results to 5 additional popular driving datasets to demonstrate its generalizability on common urban scenes.

cs.CV↗

arXiv2025

In-situ three-dimensional strain engineering of solid-state quantum emitters in photonic structures towards scalable quantum networks

Solid-state quantum emitters are pivotal for modern photonic quantum technology, yet their inherent spectral inhomogeneity imposes a critical challenge in pursuing scalable quantum network. Here, we develop a cryogenic-compatible strain-engineering platform based on a polydimethylsiloxane (PDMS) stamp that is not obviously working properly at cryogenic temperature. In-situ three-dimensional (3D) strain control is achieved for quantum dots (QDs) embedded in photonic nanostructures. The compliant PDMS enables independent tuning of emission energy and elimination of fine structure splitting (FSS) of single QDs, as demonstrated by a 7 meV spectral shift with a near-vanishing FSS in circular Bragg resonators and an unprecedented 15 meV tuning range in the micropillar. The PDMS-based 3D strain-engineering platform, compatible with diverse photonic structures at cryogenic temperature, provides a powerful and versatile tool for exploring fundamental strain-related physics and advancing integrated photonic quantum technology.

physics.optics↗

arXiv2025

SEAP: Training-free Sparse Expert Activation Pruning Unlock the Brainpower of Large Language Models

Large Language Models have achieved remarkable success across various natural language processing tasks, yet their high computational cost during inference remains a major bottleneck. This paper introduces Sparse Expert Activation Pruning (SEAP), a training-free pruning method that selectively retains task-relevant parameters to reduce inference overhead. Inspired by the clustering patterns of hidden states and activations in LLMs, SEAP identifies task-specific expert activation patterns and prunes the model while preserving task performance and enhancing computational efficiency. Experimental results demonstrate that SEAP significantly reduces computational overhead while maintaining competitive accuracy. Notably, at 50% pruning, SEAP surpasses both WandA and FLAP by over 20%, and at 20% pruning, it incurs only a 2.2% performance drop compared to the dense model. These findings highlight SEAP's scalability and effectiveness, making it a promising approach for optimizing large-scale LLMs.

cs.CL↗

arXiv2025

SampleMix: A Sample-wise Pre-training Data Mixing Strategey by Coordinating Data Quality and Diversity

Existing pretraining data mixing methods for large language models (LLMs) typically follow a domain-wise methodology, a top-down process that first determines domain weights and then performs uniform data sampling across each domain. However, these approaches neglect significant inter-domain overlaps and commonalities, failing to control the global diversity of the constructed training dataset. Further, uniform sampling within domains ignores fine-grained sample-specific features, potentially leading to suboptimal data distribution. To address these shortcomings, we propose a novel sample-wise data mixture approach based on a bottom-up paradigm. This method performs global cross-domain sampling by systematically evaluating the quality and diversity of each sample, thereby dynamically determining the optimal domain distribution. Comprehensive experiments across multiple downstream tasks and perplexity assessments demonstrate that SampleMix surpasses existing domain-based methods. Meanwhile, SampleMix requires 1.4x to 2.1x training steps to achieves the baselines' performance, highlighting the substantial potential of SampleMix to optimize pre-training data.

cs.CL↗

arXiv2025

SurveyX: Academic Survey Automation via Large Language Models

Large Language Models (LLMs) have demonstrated exceptional comprehension capabilities and a vast knowledge base, suggesting that LLMs can serve as efficient tools for automated survey generation. However, recent research related to automated survey generation remains constrained by some critical limitations like finite context window, lack of in-depth content discussion, and absence of systematic evaluation frameworks. Inspired by human writing processes, we propose SurveyX, an efficient and organized system for automated survey generation that decomposes the survey composing process into two phases: the Preparation and Generation phases. By innovatively introducing online reference retrieval, a pre-processing method called AttributeTree, and a re-polishing process, SurveyX significantly enhances the efficacy of survey composition. Experimental evaluation results show that SurveyX outperforms existing automated survey generation systems in content quality (0.259 improvement) and citation quality (1.76 enhancement), approaching human expert performance across multiple evaluation dimensions. Examples of surveys generated by SurveyX are available on www.surveyx.cn

cs.CL↗

arXiv2025

SafeRAG: Benchmarking Security in Retrieval-Augmented Generation of Large Language Model

The indexing-retrieval-generation paradigm of retrieval-augmented generation (RAG) has been highly successful in solving knowledge-intensive tasks by integrating external knowledge into large language models (LLMs). However, the incorporation of external and unverified knowledge increases the vulnerability of LLMs because attackers can perform attack tasks by manipulating knowledge. In this paper, we introduce a benchmark named SafeRAG designed to evaluate the RAG security. First, we classify attack tasks into silver noise, inter-context conflict, soft ad, and white Denial-of-Service. Next, we construct RAG security evaluation dataset (i.e., SafeRAG dataset) primarily manually for each task. We then utilize the SafeRAG dataset to simulate various attack scenarios that RAG may encounter. Experiments conducted on 14 representative RAG components demonstrate that RAG exhibits significant vulnerability to all attack tasks and even the most apparent attack task can easily bypass existing retrievers, filters, or advanced LLMs, resulting in the degradation of RAG service quality. Code is available at: https://github.com/IAAR-Shanghai/SafeRAG.

cs.CR↗

arXiv2024

STORM: Spatio-Temporal Reconstruction Model for Large-Scale Outdoor Scenes

We present STORM, a spatio-temporal reconstruction model designed for reconstructing dynamic outdoor scenes from sparse observations. Existing dynamic reconstruction methods often rely on per-scene optimization, dense observations across space and time, and strong motion supervision, resulting in lengthy optimization times, limited generalization to novel views or scenes, and degenerated quality caused by noisy pseudo-labels for dynamics. To address these challenges, STORM leverages a data-driven Transformer architecture that directly infers dynamic 3D scene representations--parameterized by 3D Gaussians and their velocities--in a single forward pass. Our key design is to aggregate 3D Gaussians from all frames using self-supervised scene flows, transforming them to the target timestep to enable complete (i.e., "amodal") reconstructions from arbitrary viewpoints at any moment in time. As an emergent property, STORM automatically captures dynamic instances and generates high-quality masks using only reconstruction losses. Extensive experiments on public datasets show that STORM achieves precise dynamic scene reconstruction, surpassing state-of-the-art per-scene optimization methods (+4.3 to 6.6 PSNR) and existing feed-forward approaches (+2.1 to 4.7 PSNR) in dynamic regions. STORM reconstructs large-scale outdoor scenes in 200ms, supports real-time rendering, and outperforms competitors in scene flow estimation, improving 3D EPE by 0.422m and Acc5 by 28.02%. Beyond reconstruction, we showcase four additional applications of our model, illustrating the potential of self-supervised learning for broader dynamic scene understanding.

cs.CV↗

arXiv2024

Map Optical Properties to Subwavelength Structures Directly via a Diffusion Model

Subwavelength photonic structures and metamaterials provide revolutionary approaches for controlling light. The inverse design methods proposed for these subwavelength structures are vital to the development of new photonic devices. However, most of the existing inverse design methods cannot realize direct mapping from optical properties to photonic structures but instead rely on forward simulation methods to perform iterative optimization. In this work, we exploit the powerful generative abilities of artificial intelligence (AI) and propose a practical inverse design method based on latent diffusion models. Our method maps directly the optical properties to structures without the requirement of forward simulation and iterative optimization. Here, the given optical properties can work as "prompts" and guide the constructed model to correctly "draw" the required photonic structures. Experiments show that our direct mapping-based inverse design method can generate subwavelength photonic structures at high fidelity while following the given optical properties. This may change the method used for optical design and greatly accelerate the research on new photonic devices.

physics.optics↗