arXiv subjects

Linfeng Zhang

Publications and source records attributed to Linfeng Zhang.

At least 145 records · Page 8Linked to original sources

arXiv2025

D2Pruner: Debiased Importance and Structural Diversity for MLLM Token Pruning

Processing long visual token sequences poses a significant computational burden on Multimodal Large Language Models (MLLMs). While token pruning offers a path to acceleration, we find that current methods, while adequate for general understanding, catastrophically fail on fine-grained localization tasks. We attribute this failure to the inherent flaws of the two prevailing strategies: importance-based methods suffer from a strong positional bias, an inherent model artifact that distracts from semantic content, while diversity-based methods exhibit structural blindness, disregarding the user's prompt and spatial redundancy. To address this, we introduce D2Pruner, a framework that rectifies these issues by uniquely combining debiased importance with a structural pruning mechanism. Our method first secures a core set of the most critical tokens as pivots based on a debiased attention score. It then performs a Maximal Independent Set (MIS) selection on the remaining tokens, which are modeled on a hybrid graph where edges signify spatial proximity and semantic similarity. This process iteratively preserves the most important and available token while removing its neighbors, ensuring that the supplementary tokens are chosen to maximize importance and diversity. Extensive experiments demonstrate that D2Pruner has exceptional efficiency and fidelity. Applied to LLaVA-1.5-7B for general understanding tasks, it reduces FLOPs by 74.2\% while retaining 99.2\% of its original performance. Furthermore, in challenging localization benchmarks with InternVL-2.5-8B, it maintains 85.7\% performance at a 90\% token reduction rate, marking a significant advancement with up to 63. 53\% improvement over existing methods.

cs.CV↗

arXiv2025

UniLabOS: An AI-Native Operating System for Autonomous Laboratories

Autonomous laboratories promise to accelerate discovery by coupling learning algorithms with robotic experimentation, yet adoption remains limited by fragmented software that separates high-level planning from low-level execution. Here we present UniLabOS, an AI-native operating system for autonomous laboratories that bridges digital decision-making and embodied experimentation through typed, stateful abstractions and transactional safeguards. UniLabOS unifies laboratory elements via an Action/Resource/Action&Resource (A/R/A&R) model, represents laboratory structure with a dual-topology of logical ownership and physical connectivity, and reconciles digital state with material motion using a transactional CRUTD protocol. Built on a distributed edge-cloud architecture with decentralized discovery, UniLabOS enables protocol mobility across reconfigurable topologies while supporting human-in-the-loop governance. We demonstrate the system in four real-world settings -- a liquid-handling workstation, a modular organic synthesis platform, a distributed electrolyte foundry, and a decentralized computation-intensive closed-loop system -- showing robust orchestration across heterogeneous instruments and multi-node coordination. UniLabOS establishes a scalable foundation for agent-ready, reproducible, and provenance-aware autonomous experimentation.

cs.CE↗

arXiv2025

Bohrium + SciMaster: Building the Infrastructure and Ecosystem for Agentic Science at Scale

AI agents are emerging as a practical way to run multi-step scientific workflows that interleave reasoning with tool use and verification, pointing to a shift from isolated AI-assisted steps toward \emph{agentic science at scale}. This shift is increasingly feasible, as scientific tools and models can be invoked through stable interfaces and verified with recorded execution traces, and increasingly necessary, as AI accelerates scientific output and stresses the peer-review and publication pipeline, raising the bar for traceability and credible evaluation. However, scaling agentic science remains difficult: workflows are hard to observe and reproduce; many tools and laboratory systems are not agent-ready; execution is hard to trace and govern; and prototype AI Scientist systems are often bespoke, limiting reuse and systematic improvement from real workflow signals. We argue that scaling agentic science requires an infrastructure-and-ecosystem approach, instantiated in Bohrium+SciMaster. Bohrium acts as a managed, traceable hub for AI4S assets -- akin to a HuggingFace of AI for Science -- that turns diverse scientific data, software, compute, and laboratory systems into agent-ready capabilities. SciMaster orchestrates these capabilities into long-horizon scientific workflows, on which scientific agents can be composed and executed. Between infrastructure and orchestration, a \emph{scientific intelligence substrate} organizes reusable models, knowledge, and components into executable building blocks for workflow reasoning and action, enabling composition, auditability, and improvement through use. We demonstrate this stack with eleven representative master agents in real workflows, achieving orders-of-magnitude reductions in end-to-end scientific cycle time and generating execution-grounded signals from real workloads at multi-million scale.

cs.AI↗

arXiv2025

IPCV: Information-Preserving Compression for MLLM Visual Encoders

Multimodal Large Language Models (MLLMs) deliver strong vision-language performance but at high computational cost, driven by numerous visual tokens processed by the Vision Transformer (ViT) encoder. Existing token pruning strategies are inadequate: LLM-stage token pruning overlooks the ViT's overhead, while conventional ViT token pruning, without language guidance, risks discarding textually critical visual cues and introduces feature distortions amplified by the ViT's bidirectional attention. To meet these challenges, we propose IPCV, a training-free, information-preserving compression framework for MLLM visual encoders. IPCV enables aggressive token pruning inside the ViT via Neighbor-Guided Reconstruction (NGR) that temporarily reconstructs pruned tokens to participate in attention with minimal overhead, then fully restores them before passing to the LLM. Besides, we introduce Attention Stabilization (AS) to further alleviate the negative influence from token pruning by approximating the K/V of pruned tokens. It can be directly applied to previous LLM-side token pruning methods to enhance their performance. Extensive experiments show that IPCV substantially reduces end-to-end computation and outperforms state-of-the-art training-free token compression methods across diverse image and video benchmarks. Our code is available at https://github.com/Perkzi/IPCV.

cs.CV↗

arXiv2025

SDAR-VL: Stable and Efficient Block-wise Diffusion for Vision-Language Understanding

Block-wise discrete diffusion offers an attractive balance between parallel generation and causal dependency modeling, making it a promising backbone for vision-language modeling. However, its practical adoption has been limited by high training cost, slow convergence, and instability, which have so far kept it behind strong autoregressive (AR) baselines. We present \textbf{SDAR-VL}, the first systematic application of block-wise discrete diffusion to large-scale vision-language understanding (VLU), together with an \emph{integrated framework for efficient and stable training}. This framework unifies three components: (1) \textbf{Asynchronous Block-wise Noise Scheduling} to diversify supervision within each batch; (2) \textbf{Effective Mask Ratio Scaling} for unbiased loss normalization under stochastic masking; and (3) a \textbf{Progressive Beta Noise Curriculum} that increases effective mask coverage while preserving corruption diversity. Experiments on 21 single-image, multi-image, and video benchmarks show that SDAR-VL consistently improves \emph{training efficiency}, \emph{convergence stability}, and \emph{task performance} over conventional block diffusion. On this evaluation suite, SDAR-VL sets a new state of the art among diffusion-based vision-language models and, under matched settings, matches or surpasses strong AR baselines such as LLaVA-OneVision as well as the global diffusion baseline LLaDA-V, establishing block-wise diffusion as a practical backbone for VLU.

cs.CV↗

arXiv2025

ViCO: A Training Strategy towards Semantic Aware Dynamic High-Resolution

Existing Multimodal Large Language Models (MLLMs) suffer from increased inference costs due to the additional vision tokens introduced by image inputs. In this work, we propose Visual Consistency Learning (ViCO), a novel training algorithm that enables the model to represent images of varying semantic complexities using different numbers of vision tokens. The key idea behind our method is to employ multiple MLP connectors, each with a different image compression ratio, to downsample the vision tokens based on the semantic complexity of the image. During training, we minimize the KL divergence between the responses conditioned on different MLP connectors. At inference time, we introduce an image router, termed Visual Resolution Router (ViR), that automatically selects the appropriate compression rate for each image patch. Compared with existing dynamic high-resolution strategies, which adjust the number of visual tokens based on image resolutions, our method dynamically adapts the number of visual tokens according to semantic complexity. Experimental results demonstrate that our method can reduce the number of vision tokens by up to 50% while maintaining the model's perception, reasoning, and OCR capabilities. We hope this work will contribute to the development of more efficient MLLMs. The code and models will be released to facilitate future research.

cs.CV↗

arXiv2025

Integrating Deep-Learning-Based Magnetic Model and Non-Collinear Spin-Constrained Method: Methodology, Implementation and Application

We propose a non-collinear spin-constrained method that generates training data for deep-learning-based magnetic model, which provides a powerful tool for studying complex magnetic phenomena that requires large-scale simulations at the atomic level. First, we propose a basis-independent projection method for calculating atomic magnetic moments by applying a radial truncation to numerical atomic orbitals. A double-loop Lagrange multiplier method is utilized to ensure the satisfaction of constraint conditions while achieving accurate magnetic torque. The method is implemented in ABACUS with both plane wave basis and numerical atomic orbital basis. We benchmark the iron (Fe) systems and analyze differences from calculations with the plane wave basis and numerical atomic orbitals basis in describing magnetic energy barriers. Based on an automated workflow composed of first-principles calculations, magnetic model, active learning, and dynamics simulation, more than 30,000 first-principles data with the information of magnetic torque are generated to train a deep-learning-based magnetic model DeePSPIN for the Fe system. By utilizing the model in large-scale molecular dynamics simulations, we successfully predict Curie temperatures of alpha-Fe close to experimental values.

cond-mat.mtrl-sci↗

arXiv2025

EVCtrl: Efficient Control Adapter for Visual Generation

Visual generation includes both image and video generation, training probabilistic models to create coherent, diverse, and semantically faithful content from scratch. While early research focused on unconditional sampling, practitioners now demand controllable generation that allows precise specification of layout, pose, motion, or style. While ControlNet grants precise spatial-temporal control, its auxiliary branch markedly increases latency and introduces redundant computation in both uncontrolled regions and denoising steps, especially for video. To address this problem, we introduce EVCtrl, a lightweight, plug-and-play control adapter that slashes overhead without retraining the model. Specifically, we propose a spatio-temporal dual caching strategy for sparse control information. For spatial redundancy, we first profile how each layer of DiT-ControlNet responds to fine-grained control, then partition the network into global and local functional zones. A locality-aware cache focuses computation on the local zones that truly need the control signal, skipping the bulk of redundant computation in global regions. For temporal redundancy, we selectively omit unnecessary denoising steps to improve efficiency. Extensive experiments on CogVideo-Controlnet, Wan2.1-Controlnet, and Flux demonstrate that our method is effective in image and video control generation without the need for training. For example, it achieves 2.16 and 2.05 times speedups on CogVideo-Controlnet and Wan2.1-Controlnet, respectively, with almost no degradation in generation quality.Codes are available in the supplementary materials.

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

VideoCompressa: Data-Efficient Video Understanding via Joint Temporal Compression and Spatial Reconstruction

The scalability of video understanding models is increasingly limited by the prohibitive storage and computational costs of large-scale video datasets. While data synthesis has improved data efficiency in the image domain, its extension to video remains challenging due to pervasive temporal redundancy and complex spatiotemporal dynamics. In this work, we uncover a critical insight: the primary source of inefficiency in video datasets is not inter-sample redundancy, but intra-sample frame-level redundancy. To leverage this insight, we introduce VideoCompressa, a novel framework for video data synthesis that reframes the problem as dynamic latent compression. Specifically, VideoCompressa jointly optimizes a differentiable keyframe selector-implemented as a lightweight ConvNet with Gumbel-Softmax sampling-to identify the most informative frames, and a pretrained, frozen Variational Autoencoder (VAE) to compress these frames into compact, semantically rich latent codes. These latent representations are then fed into a compression network, enabling end-to-end backpropagation. Crucially, the keyframe selector and synthetic latent codes are co-optimized to maximize retention of task-relevant information. Experiments show that our method achieves unprecedented data efficiency: on UCF101 with ConvNets, VideoCompressa surpasses full-data training by 2.34\% points using only 0.13\% of the original data, with over 5800x speedup compared to traditional synthesis method. Moreover, when fine-tuning Qwen2.5-7B-VL on HMDB51, VideoCompressa matches full-data performance using just 0.41\% of the training data-outperforming zero-shot baseline by 10.61\%.

cs.CV↗

arXiv2025

Video Compression Commander: Plug-and-Play Inference Acceleration for Video Large Language Models

Video large language models (VideoLLM) excel at video understanding, but face efficiency challenges due to the quadratic complexity of abundant visual tokens. Our systematic analysis of token compression methods for VideoLLMs reveals two critical issues: (i) overlooking distinctive visual signals across frames, leading to information loss; (ii) suffering from implementation constraints, causing incompatibility with modern architectures or efficient operators. To address these challenges, we distill three design principles for VideoLLM token compression and propose a plug-and-play inference acceleration framework "Video Compression Commander" (VidCom2). By quantifying each frame's uniqueness, VidCom2 adaptively adjusts compression intensity across frames, effectively preserving essential information while reducing redundancy in video sequences. Extensive experiments across various VideoLLMs and benchmarks demonstrate the superior performance and efficiency of our VidCom2. With only 25% visual tokens, VidCom2 achieves 99.6% of the original performance on LLaVA-OV while reducing 70.8% of the LLM generation latency. Notably, our Frame Compression Adjustment strategy is compatible with other token compression methods to further improve their performance. Our code is available at https://github.com/xuyang-liu16/VidCom2.

cs.CV↗

arXiv2025

UNSEEN: Enhancing Dataset Pruning from a Generalization Perspective

The growing scale of datasets in deep learning has introduced significant computational challenges. Dataset pruning addresses this challenge by constructing a compact but informative coreset from the full dataset with comparable performance. Previous approaches typically establish scoring metrics based on specific criteria to identify representative samples. However, these methods predominantly rely on sample scores obtained from the model's performance during the training (i.e., fitting) phase. As scoring models achieve near-optimal performance on training data, such fitting-centric approaches induce a dense distribution of sample scores within a narrow numerical range. This concentration reduces the distinction between samples and hinders effective selection. To address this challenge, we conduct dataset pruning from the perspective of generalization, i.e., scoring samples based on models not exposed to them during training. We propose a plug-and-play framework, UNSEEN, which can be integrated into existing dataset pruning methods. Additionally, conventional score-based methods are single-step and rely on models trained solely on the complete dataset, providing limited perspective on the importance of samples. To address this limitation, we scale UNSEEN to multi-step scenarios and propose an incremental selection technique through scoring models trained on varying coresets, and optimize the quality of the coreset dynamically. Extensive experiments demonstrate that our method significantly outperforms existing state-of-the-art (SOTA) methods on CIFAR-10, CIFAR-100, and ImageNet-1K. Notably, on ImageNet-1K, UNSEEN achieves lossless performance while reducing training data by 30\%.

cs.CV↗

arXiv2025

InterCLIP-MEP: Interactive CLIP and Memory-Enhanced Predictor for Multi-modal Sarcasm Detection

Sarcasm in social media, frequently conveyed through the interplay of text and images, presents significant challenges for sentiment analysis and intention mining. Existing multi-modal sarcasm detection approaches have been shown to excessively depend on superficial cues within the textual modality, exhibiting limited capability to accurately discern sarcasm through subtle text-image interactions. To address this limitation, a novel framework, InterCLIP-MEP, is proposed. This framework integrates Interactive CLIP (InterCLIP), which employs an efficient training strategy to derive enriched cross-modal representations by embedding inter-modal information directly into each encoder, while using approximately 20.6$\times$ fewer trainable parameters compared with existing state-of-the-art (SOTA) methods. Furthermore, a Memory-Enhanced Predictor (MEP) is introduced, featuring a dynamic dual-channel memory mechanism that captures and retains valuable knowledge from test samples during inference, serving as a non-parametric classifier to enhance sarcasm detection robustness. Extensive experiments on MMSD, MMSD2.0, and DocMSU show that InterCLIP-MEP achieves SOTA performance, specifically improving accuracy by 1.08% and F1 score by 1.51% on MMSD2.0. Under distributional shift evaluation, it attains 73.96% accuracy, exceeding its memory-free variant by nearly 10% and the previous SOTA by over 15%, demonstrating superior stability and adaptability. The implementation of InterCLIP-MEP is publicly available at https://github.com/CoderChen01/InterCLIP-MEP.

cs.CL↗

arXiv2025

ImagebindDC: Compressing Multi-modal Data with Imagebind-based Condensation

Data condensation techniques aim to synthesize a compact dataset from a larger one to enable efficient model training, yet while successful in unimodal settings, they often fail in multimodal scenarios where preserving intricate inter-modal dependencies is crucial. To address this, we introduce ImageBindDC, a novel data condensation framework operating within the unified feature space of ImageBind. Our approach moves beyond conventional distribution-matching by employing a powerful Characteristic Function (CF) loss, which operates in the Fourier domain to facilitate a more precise statistical alignment via exact infinite moment matching. We design our objective to enforce three critical levels of distributional consistency: (i) uni-modal alignment, which matches the statistical properties of synthetic and real data within each modality; (ii) cross-modal alignment, which preserves pairwise semantics by matching the distributions of hybrid real-synthetic data pairs; and (iii) joint-modal alignment, which captures the complete multivariate data structure by aligning the joint distribution of real data pairs with their synthetic counterparts. Extensive experiments highlight the effectiveness of ImageBindDC: on the NYU-v2 dataset, a model trained on just 5 condensed datapoints per class achieves lossless performance comparable to one trained on the full dataset, achieving a new state-of-the-art with an 8.2\% absolute improvement over the previous best method and more than 4$\times$ less condensation time.

cs.CV↗

arXiv2025

SageLM: A Multi-aspect and Explainable Large Language Model for Speech Judgement

Speech-to-Speech (S2S) Large Language Models (LLMs) are foundational to natural human-computer interaction, enabling end-to-end spoken dialogue systems. However, evaluating these models remains a fundamental challenge. We propose \texttt{SageLM}, an end-to-end, multi-aspect, and explainable speech LLM for comprehensive S2S LLMs evaluation. First, unlike cascaded approaches that disregard acoustic features, SageLM jointly assesses both semantic and acoustic dimensions. Second, it leverages rationale-based supervision to enhance explainability and guide model learning, achieving superior alignment with evaluation outcomes compared to rule-based reinforcement learning methods. Third, we introduce \textit{SpeechFeedback}, a synthetic preference dataset, and employ a two-stage training paradigm to mitigate the scarcity of speech preference data. Trained on both semantic and acoustic dimensions, SageLM achieves an 82.79\% agreement rate with human evaluators, outperforming cascaded and SLM-based baselines by at least 7.42\% and 26.20\%, respectively.

cs.CL↗

arXiv2025

Multimodal Spatial Reasoning in the Large Model Era: A Survey and Benchmarks

Humans possess spatial reasoning abilities that enable them to understand spaces through multimodal observations, such as vision and sound. Large multimodal reasoning models extend these abilities by learning to perceive and reason, showing promising performance across diverse spatial tasks. However, systematic reviews and publicly available benchmarks for these models remain limited. In this survey, we provide a comprehensive review of multimodal spatial reasoning tasks with large models, categorizing recent progress in multimodal large language models (MLLMs) and introducing open benchmarks for evaluation. We begin by outlining general spatial reasoning, focusing on post-training techniques, explainability, and architecture. Beyond classical 2D tasks, we examine spatial relationship reasoning, scene and layout understanding, as well as visual question answering and grounding in 3D space. We also review advances in embodied AI, including vision-language navigation and action models. Additionally, we consider emerging modalities such as audio and egocentric video, which contribute to novel spatial understanding through new sensors. We believe this survey establishes a solid foundation and offers insights into the growing field of multimodal spatial reasoning. Updated information about this survey, codes and implementation of the open benchmarks can be found at https://github.com/zhengxuJosh/Awesome-Spatial-Reasoning.

cs.CV↗

arXiv2025

A Survey on Cache Methods in Diffusion Models: Toward Efficient Multi-Modal Generation

Diffusion Models have become a cornerstone of modern generative AI for their exceptional generation quality and controllability. However, their inherent \textit{multi-step iterations} and \textit{complex backbone networks} lead to prohibitive computational overhead and generation latency, forming a major bottleneck for real-time applications. Although existing acceleration techniques have made progress, they still face challenges such as limited applicability, high training costs, or quality degradation. Against this backdrop, \textbf{Diffusion Caching} offers a promising training-free, architecture-agnostic, and efficient inference paradigm. Its core mechanism identifies and reuses intrinsic computational redundancies in the diffusion process. By enabling feature-level cross-step reuse and inter-layer scheduling, it reduces computation without modifying model parameters. This paper systematically reviews the theoretical foundations and evolution of Diffusion Caching and proposes a unified framework for its classification and analysis. Through comparative analysis of representative methods, we show that Diffusion Caching evolves from \textit{static reuse} to \textit{dynamic prediction}. This trend enhances caching flexibility across diverse tasks and enables integration with other acceleration techniques such as sampling optimization and model distillation, paving the way for a unified, efficient inference framework for future multimodal and interactive applications. We argue that this paradigm will become a key enabler of real-time and efficient generative AI, injecting new vitality into both theory and practice of \textit{Efficient Generative Intelligence}.

cs.LG↗

arXiv2025

Diffusion LLM with Native Variable Generation Lengths: Let [EOS] Lead the Way

Diffusion-based large language models (dLLMs) have exhibited substantial potential for parallel text generation, which may enable more efficient generation compared to autoregressive models. However, current dLLMs suffer from fixed generation lengths, which indicates the generation lengths of dLLMs have to be determined before decoding as a hyper-parameter, leading to issues in efficiency and flexibility. To solve these problems, in this work, we propose to train a diffusion LLM with native variable generation lengths, abbreviated as dLLM-Var. Concretely, we aim to train a model to accurately predict the [EOS] token in the generated text, which makes a dLLM be able to natively infer in a block diffusion manner, while still maintaining the ability of global bi-directional (full) attention and high parallelism. Experiments on standard benchmarks demonstrate that our method achieves a 30.1x speedup over traditional dLLM inference paradigms and a 2.4x speedup relative to autoregressive models such as Qwen and Llama. Our method achieves higher accuracy and faster inference, elevating dLLMs beyond mere academic novelty and supporting their practical use in real-world applications. Codes and models have been released.

cs.CL↗