arXiv subjects

Pan Zhang

Publications and source records attributed to Pan Zhang.

At least 73 records · Page 4Linked to original sources

arXiv2025

OVO-Bench: How Far is Your Video-LLMs from Real-World Online Video Understanding?

Temporal Awareness, the ability to reason dynamically based on the timestamp when a question is raised, is the key distinction between offline and online video LLMs. Unlike offline models, which rely on complete videos for static, post hoc analysis, online models process video streams incrementally and dynamically adapt their responses based on the timestamp at which the question is posed. Despite its significance, temporal awareness has not been adequately evaluated in existing benchmarks. To fill this gap, we present OVO-Bench (Online-VideO-Benchmark), a novel video benchmark that emphasizes the importance of timestamps for advanced online video understanding capability benchmarking. OVO-Bench evaluates the ability of video LLMs to reason and respond to events occurring at specific timestamps under three distinct scenarios: (1) Backward tracing: trace back to past events to answer the question. (2) Real-time understanding: understand and respond to events as they unfold at the current timestamp. (3) Forward active responding: delay the response until sufficient future information becomes available to answer the question accurately. OVO-Bench comprises 12 tasks, featuring 644 unique videos and approximately human-curated 2,800 fine-grained meta-annotations with precise timestamps. We combine automated generation pipelines with human curation. With these high-quality samples, we further developed an evaluation pipeline to systematically query video LLMs along the video timeline. Evaluations of nine Video-LLMs reveal that, despite advancements on traditional benchmarks, current models struggle with online video understanding, showing a significant gap compared to human agents. We hope OVO-Bench will drive progress in video LLMs and inspire future research in online video reasoning. Our benchmark and code can be accessed at https://github.com/JoeLeelyf/OVO-Bench.

cs.CV↗

arXiv2025

Generative Decoding for Quantum Error-correcting Codes

Efficient and accurate decoding of quantum error-correcting codes is essential for fault-tolerant quantum computation, however, it is challenging due to the degeneracy of errors, the complex code topology, and the large space for logical operators in high-rate codes. In this work, we propose a decoding algorithm utilizing generative modeling in machine learning. We employ autoregressive neural networks to learn the joint probability of logical operators and syndromes in an unsupervised manner, eliminating the need for labeled training data. The learned model can approximately perform maximum likelihood decoding by directly generating the most likely logical operators for $k$ logical qubits with $\mathcal O(2k)$ computational complexity. Thus, it is particularly efficient for decoding high-rate codes with many logical qubits. The proposed approach is general and applies to a wide spectrum of quantum error-correcting codes including surface codes and quantum low-density parity-check codes (qLDPC), under noise models ranging from code capacity noise to circuit level noise. We conducted extensive numerical experiments to demonstrate that our approach achieves significantly higher decoding accuracy compared to the minimum weight perfect matching and belief propagation with ordered statistics on the surface codes and high-rate quantum low-density parity-check codes. Our approach highlights generative artificial intelligence as a potential solution for the real-time decoding of realistic and high-rate quantum error correction codes.

quant-ph↗

arXiv2025

Light-A-Video: Training-free Video Relighting via Progressive Light Fusion

Recent advancements in image relighting models, driven by large-scale datasets and pre-trained diffusion models, have enabled the imposition of consistent lighting. However, video relighting still lags, primarily due to the excessive training costs and the scarcity of diverse, high-quality video relighting datasets. A simple application of image relighting models on a frame-by-frame basis leads to several issues: lighting source inconsistency and relighted appearance inconsistency, resulting in flickers in the generated videos. In this work, we propose Light-A-Video, a training-free approach to achieve temporally smooth video relighting. Adapted from image relighting models, Light-A-Video introduces two key techniques to enhance lighting consistency. First, we design a Consistent Light Attention (CLA) module, which enhances cross-frame interactions within the self-attention layers of the image relight model to stabilize the generation of the background lighting source. Second, leveraging the physical principle of light transport independence, we apply linear blending between the source video's appearance and the relighted appearance, using a Progressive Light Fusion (PLF) strategy to ensure smooth temporal transitions in illumination. Experiments show that Light-A-Video improves the temporal consistency of relighted video while maintaining the relighted image quality, ensuring coherent lighting transitions across frames. Project page: https://bujiazi.github.io/light-a-video.github.io/.

cs.CV↗

arXiv2025

Rapid morphology characterization of two-dimensional TMDs and lateral heterostructures based on deep learning

Two-dimensional (2D) materials and heterostructures exhibit unique physical properties, necessitating efficient and accurate characterization methods. Leveraging advancements in artificial intelligence, we introduce a deep learning-based method for efficiently characterizing heterostructures and 2D materials, specifically MoS2-MoSe2 lateral heterostructures and MoS2 flakes with varying shapes and thicknesses. By utilizing YOLO models, we achieve an accuracy rate of over 94.67% in identifying these materials. Additionally, we explore the application of transfer learning across different materials, which further enhances model performance. This model exhibits robust generalization and anti-interference ability, ensuring reliable results in diverse scenarios. To facilitate practical use, we have developed an application that enables real-time analysis directly from optical microscope images, making the process significantly faster and more cost-effective than traditional methods. This deep learning-driven approach represents a promising tool for the rapid and accurate characterization of 2D materials, opening new avenues for research and development in material science.

cs.LG↗

arXiv2025

ByTheWay: Boost Your Text-to-Video Generation Model to Higher Quality in a Training-free Way

The text-to-video (T2V) generation models, offering convenient visual creation, have recently garnered increasing attention. Despite their substantial potential, the generated videos may present artifacts, including structural implausibility, temporal inconsistency, and a lack of motion, often resulting in near-static video. In this work, we have identified a correlation between the disparity of temporal attention maps across different blocks and the occurrence of temporal inconsistencies. Additionally, we have observed that the energy contained within the temporal attention maps is directly related to the magnitude of motion amplitude in the generated videos. Based on these observations, we present ByTheWay, a training-free method to improve the quality of text-to-video generation without introducing additional parameters, augmenting memory or sampling time. Specifically, ByTheWay is composed of two principal components: 1) Temporal Self-Guidance improves the structural plausibility and temporal consistency of generated videos by reducing the disparity between the temporal attention maps across various decoder blocks. 2) Fourier-based Motion Enhancement enhances the magnitude and richness of motion by amplifying the energy of the map. Extensive experiments demonstrate that ByTheWay significantly improves the quality of text-to-video generation with negligible additional cost.

cs.CV↗

arXiv2025

PyramidDrop: Accelerating Your Large Vision-Language Models via Pyramid Visual Redundancy Reduction

In large vision-language models (LVLMs), images serve as inputs that carry a wealth of information. As the idiom "A picture is worth a thousand words" implies, representing a single image in current LVLMs can require hundreds or even thousands of tokens. This results in significant computational costs, which grow quadratically as input image resolution increases, thereby severely impacting the efficiency of both training and inference. Previous approaches have attempted to reduce the number of image tokens either before or within the early layers of LVLMs. However, these strategies inevitably result in the loss of crucial image information, ultimately diminishing model performance. To address this challenge, we conduct an empirical study revealing that all visual tokens are necessary for LVLMs in the shallow layers, and token redundancy progressively increases in the deeper layers of the model. To this end, we propose PyramidDrop, a visual redundancy reduction strategy for LVLMs to boost their efficiency in both training and inference with neglectable performance loss. Specifically, we partition the LVLM into several stages and drop part of the image tokens at the end of each stage with a pre-defined ratio, creating pyramid-like visual tokens across model layers. The dropping is based on a lightweight similarity calculation with a negligible time overhead. Extensive experiments demonstrate that PyramidDrop can achieve a 40% training time and 55% inference FLOPs acceleration of LLaVA-NeXT with comparable performance. Besides, the PyramidDrop could also serve as a plug-and-play strategy for inference acceleration without training, with better performance and lower inference cost than counterparts. Code is available at https://github.com/Cooperx521/PyramidDrop.

cs.CV↗

arXiv2025

BoostStep: Boosting mathematical capability of Large Language Models via improved single-step reasoning

Large language models (LLMs) have demonstrated impressive ability in solving complex mathematical problems with multi-step reasoning and can be further enhanced with well-designed in-context learning (ICL) examples. However, this potential is often constrained by two major challenges in ICL: granularity mismatch and irrelevant information. We observe that while LLMs excel at decomposing mathematical problems, they often struggle with reasoning errors in fine-grained steps. Moreover, ICL examples retrieved at the question level may omit critical steps or even mislead the model with irrelevant details. To address this issue, we propose BoostStep, a method that enhances reasoning accuracy through step-aligned ICL, a novel mechanism that carefully aligns retrieved reference steps with the corresponding reasoning steps. Additionally, BoostStep incorporates an effective "first-try" strategy to deliver exemplars highly relevant to the current state of reasoning. BoostStep is a flexible and powerful method that integrates seamlessly with chain-of-thought (CoT) and tree search algorithms, refining both candidate selection and decision-making. Empirical results show that BoostStep improves GPT-4o's CoT performance by 4.6% across mathematical benchmarks, significantly surpassing traditional few-shot learning's 1.2%. Moreover, it can achieve an additional 7.5\% gain combined with tree search. Surprisingly, it enhances state-of-the-art LLMs to solve challenging math problems using simpler examples. It improves DeepSeek-R1-671B's performance on AIME by 2.2%, leveraging simple examples only from the MATH dataset.

cs.CL↗

arXiv2025

Efficient Optimization of Variational Autoregressive Networks with Natural Gradient

Estimating free energy is a fundamental problem in statistical mechanics. Recently, machine-learning-based methods, particularly the variational autoregressive networks (VANs) have been proposed to minimize variational free energy and to approximate the Boltzmann distribution. VAN enjoys notable advantages, including the exact computation of the normalized joint distribution and fast sampling, which are critical features often missing in Markov chain Monte Carlo algorithms. However, VAN also faces significant computational challenges. These include difficulties in the optimization of variational free energy in a complicated parameter space and slow convergence of learning. In this work, we introduce an optimization technique based on natural gradients to the VAN framework, namely ng-VAN, to enhance the learning efficiency and accuracy of the conventional VAN. The method has computational complexity cubic in the batch size rather than in the number of model parameters, hence it can be efficiently implemented for a large VAN model. We carried out extensive numerical experiments on the Sherrington-Kirkpatrick model, spin glasses on random graphs, and the two-dimensional Ising model. Our results indicate that compared with the conventional VAN, ng-VAN significantly improves the accuracy in estimating free energy and converges much faster with shorter learning time. This allows extending the VAN framework's applicability to challenging statistical mechanics problems that were previously not accessible.

cond-mat.stat-mech↗

arXiv2025

RelightVid: Temporal-Consistent Diffusion Model for Video Relighting

Diffusion models have demonstrated remarkable success in image generation and editing, with recent advancements enabling albedo-preserving image relighting. However, applying these models to video relighting remains challenging due to the lack of paired video relighting datasets and the high demands for output fidelity and temporal consistency, further complicated by the inherent randomness of diffusion models. To address these challenges, we introduce RelightVid, a flexible framework for video relighting that can accept background video, text prompts, or environment maps as relighting conditions. Trained on in-the-wild videos with carefully designed illumination augmentations and rendered videos under extreme dynamic lighting, RelightVid achieves arbitrary video relighting with high temporal consistency without intrinsic decomposition while preserving the illumination priors of its image backbone.

cs.CV↗

arXiv2025

Tensor network Monte Carlo simulations for the two-dimensional random-bond Ising model

Disordered lattice spin systems are crucial in both theoretical and applied physics. However, understanding their properties poses significant challenges for Monte Carlo simulations. In this work, we investigate the two-dimensional random-bond Ising model using the recently proposed Tensor Network Monte Carlo (TNMC) method. This method generates biased samples from conditional probabilities computed via tensor network contractions and corrects the bias using the Metropolis scheme. Consequently, the proposals provided by tensor networks function as block updates for Monte Carlo simulations. Through extensive numerical experiments, we demonstrate that TNMC simulations can be performed on lattices as large as $1024\times 1024$ spins with moderate computational resources, a substantial increase from the previous maximum size of $64\times 64$ in MCMC. Notably, we observe an almost complete absence of critical slowing down, enabling the efficient collection of unbiased samples and averaging over a large number of random realizations of bond disorders. We successfully pinpoint the multi-critical point along the Nishimori line with significant precision and accurately determined the bulk and surface critical exponents. Our findings suggest that TNMC is a highly efficient algorithm for exploring disordered and frustrated systems in two dimensions.

cond-mat.stat-mech↗

arXiv2025

Exact Decoding of Repetition Code under Circuit Level Noise

Repetition code forms a fundamental basis for quantum error correction experiments. To date, it stands as the sole code that has achieved large distances and extremely low error rates. Its applications span the spectrum of evaluating hardware limitations, pinpointing hardware defects, and detecting rare events. However, current methods for decoding repetition codes under circuit level noise are suboptimal, leading to inaccurate error correction thresholds and introducing additional errors in event detection. In this work, we establish that repetition code under circuit level noise has an exact solution, and we propose an optimal maximum likelihood decoding algorithm called planar. The algorithm is based on the exact solution of the spin glass partition function on planar graphs and has polynomial computational complexity. Through extensive numerical experiments, we demonstrate that our algorithm uncovers the exact threshold for depolarizing noise and realistic superconductor SI1000 noise. Furthermore, we apply our method to analyze data from recent quantum memory experiments conducted by Google Quantum AI, revealing that part of the error floor was attributed to the decoding algorithm used by Google. Finally, we implemented the repetition code quantum memory on superconducting systems with a 72-qubit quantum chip lacking reset gates, demonstrating that even with an unknown error model, the proposed algorithm achieves a significantly lower logical error rate than the matching-based algorithm.

quant-ph↗

arXiv2025

Dispider: Enabling Video LLMs with Active Real-Time Interaction via Disentangled Perception, Decision, and Reaction

Active Real-time interaction with video LLMs introduces a new paradigm for human-computer interaction, where the model not only understands user intent but also responds while continuously processing streaming video on the fly. Unlike offline video LLMs, which analyze the entire video before answering questions, active real-time interaction requires three capabilities: 1) Perception: real-time video monitoring and interaction capturing. 2) Decision: raising proactive interaction in proper situations, 3) Reaction: continuous interaction with users. However, inherent conflicts exist among the desired capabilities. The Decision and Reaction require a contrary Perception scale and grain, and the autoregressive decoding blocks the real-time Perception and Decision during the Reaction. To unify the conflicted capabilities within a harmonious system, we present Dispider, a system that disentangles Perception, Decision, and Reaction. Dispider features a lightweight proactive streaming video processing module that tracks the video stream and identifies optimal moments for interaction. Once the interaction is triggered, an asynchronous interaction module provides detailed responses, while the processing module continues to monitor the video in the meantime. Our disentangled and asynchronous design ensures timely, contextually accurate, and computationally efficient responses, making Dispider ideal for active real-time interaction for long-duration video streams. Experiments show that Dispider not only maintains strong performance in conventional video QA tasks, but also significantly surpasses previous online models in streaming scenario responses, thereby validating the effectiveness of our architecture. The code and model are released at \url{https://github.com/Mark12Ding/Dispider}.

cs.CV↗

arXiv2024

Free-Energy Machine for Combinatorial Optimization

Finding optimal solutions to combinatorial optimization problems is pivotal in both scientific and technological domains, within academic research and industrial applications. A considerable amount of effort has been invested in the development of accelerated methods that leverage sophisticated models and harness the power of advanced computational hardware. Despite the advancements, a critical challenge persists, the dual demand for both high efficiency and broad generality in solving problems. In this work, we propose a general method, Free-Energy Machine (FEM), based on the ideas of free-energy minimization in statistical physics, combined with automatic differentiation and gradient-based optimization in machine learning. The algorithm is flexible, solving various combinatorial optimization problems using a unified framework, and is efficient, naturally utilizing massive parallel computational devices such as graph processing units (GPUs) and field-programmable gate arrays (FPGAs). We benchmark our algorithm on various problems including the maximum cut problems, balanced minimum cut problems, and maximum $k$-satisfiability problems, scaled to millions of variables, across both synthetic, real-world, and competition problem instances. The findings indicate that our algorithm not only exhibits exceptional speed but also surpasses the performance of state-of-the-art algorithms tailored for individual problems. This highlights that the interdisciplinary fusion of statistical physics and machine learning opens the door to delivering cutting-edge methodologies that will have broad implications across various scientific and industrial landscapes.

cond-mat.stat-mech↗

arXiv2024

InternLM-XComposer2.5-OmniLive: A Comprehensive Multimodal System for Long-term Streaming Video and Audio Interactions

Creating AI systems that can interact with environments over long periods, similar to human cognition, has been a longstanding research goal. Recent advancements in multimodal large language models (MLLMs) have made significant strides in open-world understanding. However, the challenge of continuous and simultaneous streaming perception, memory, and reasoning remains largely unexplored. Current MLLMs are constrained by their sequence-to-sequence architecture, which limits their ability to process inputs and generate responses simultaneously, akin to being unable to think while perceiving. Furthermore, relying on long contexts to store historical data is impractical for long-term interactions, as retaining all information becomes costly and inefficient. Therefore, rather than relying on a single foundation model to perform all functions, this project draws inspiration from the concept of the Specialized Generalist AI and introduces disentangled streaming perception, reasoning, and memory mechanisms, enabling real-time interaction with streaming video and audio input. The proposed framework InternLM-XComposer2.5-OmniLive (IXC2.5-OL) consists of three key modules: (1) Streaming Perception Module: Processes multimodal information in real-time, storing key details in memory and triggering reasoning in response to user queries. (2) Multi-modal Long Memory Module: Integrates short-term and long-term memory, compressing short-term memories into long-term ones for efficient retrieval and improved accuracy. (3) Reasoning Module: Responds to queries and executes reasoning tasks, coordinating with the perception and memory modules. This project simulates human-like cognition, enabling multimodal large language models to provide continuous and adaptive service over time.

cs.CV↗

arXiv2024

Automated Discovery of Branching Rules with Optimal Complexity for the Maximum Independent Set Problem

The branching algorithm is a fundamental technique for designing fast exponential-time algorithms to solve combinatorial optimization problems exactly. It divides the entire solution space into independent search branches using predetermined branching rules, and ignores the search on suboptimal branches to reduce the time complexity. The complexity of a branching algorithm is primarily determined by the branching rules it employs, which are often designed by human experts. In this paper, we show how to automate this process with a focus on the maximum independent set problem. The main contribution is an algorithm that efficiently generate optimal branching rules for a given sub-graph with tens of vertices. Its efficiency enables us to generate the branching rules on-the-fly, which is provably optimal and significantly reduces the number of branches compared to existing methods that rely on expert-designed branching rules. Numerical experiment on 3-regular graphs shows an average complexity of O(1.0441^n) can be achieved, better than any previous methods.

math.OC↗

arXiv2024

ASANet: Asymmetric Semantic Aligning Network for RGB and SAR image land cover classification

Synthetic Aperture Radar (SAR) images have proven to be a valuable cue for multimodal Land Cover Classification (LCC) when combined with RGB images. Most existing studies on cross-modal fusion assume that consistent feature information is necessary between the two modalities, and as a result, they construct networks without adequately addressing the unique characteristics of each modality. In this paper, we propose a novel architecture, named the Asymmetric Semantic Aligning Network (ASANet), which introduces asymmetry at the feature level to address the issue that multi-modal architectures frequently fail to fully utilize complementary features. The core of this network is the Semantic Focusing Module (SFM), which explicitly calculates differential weights for each modality to account for the modality-specific features. Furthermore, ASANet incorporates a Cascade Fusion Module (CFM), which delves deeper into channel and spatial representations to efficiently select features from the two modalities for fusion. Through the collaborative effort of these two modules, the proposed ASANet effectively learns feature correlations between the two modalities and eliminates noise caused by feature differences. Comprehensive experiments demonstrate that ASANet achieves excellent performance on three multimodal datasets. Additionally, we have established a new RGB-SAR multimodal dataset, on which our ASANet outperforms other mainstream methods with improvements ranging from 1.21% to 17.69%. The ASANet runs at 48.7 frames per second (FPS) when the input image is 256x256 pixels. The source code are available at https://github.com/whu-pzhang/ASANet

eess.IV↗

arXiv2024

MMLongBench-Doc: Benchmarking Long-context Document Understanding with Visualizations

Understanding documents with rich layouts and multi-modal components is a long-standing and practical task. Recent Large Vision-Language Models (LVLMs) have made remarkable strides in various tasks, particularly in single-page document understanding (DU). However, their abilities on long-context DU remain an open problem. This work presents MMLongBench-Doc, a long-context, multi-modal benchmark comprising 1,062 expert-annotated questions. Distinct from previous datasets, it is constructed upon 130 lengthy PDF-formatted documents with an average of 49.4 pages and 20,971 textual tokens. Towards comprehensive evaluation, answers to these questions rely on pieces of evidence from (1) different sources (text, image, chart, table, and layout structure) and (2) various locations (i.e. page number). Moreover, 33.2% of the questions are cross-page questions requiring evidence across multiple pages. 22.8% of the questions are designed to be unanswerable for detecting potential hallucinations. Experiments on 14 LVLMs demonstrate that long-context DU greatly challenges current models. Notably, the best-performing model, GPT-4o, achieves an F1 score of only 42.7%, while the second-best, GPT-4V, scores 31.4%. Furthermore, 12 LVLMs (all except GPT-4o and GPT-4V) even present worse performance than their LLM counterparts which are fed with lossy-parsed OCR documents. These results validate the necessity of future research toward more capable long-context LVLMs. Project Page: https://mayubo2333.github.io/MMLongBench-Doc

cs.CV↗

arXiv2024

Pattern Integration and Enhancement Vision Transformer for Self-Supervised Learning in Remote Sensing

Recent self-supervised learning (SSL) methods have demonstrated impressive results in learning visual representations from unlabeled remote sensing images. However, most remote sensing images predominantly consist of scenographic scenes containing multiple ground objects without explicit foreground targets, which limits the performance of existing SSL methods that focus on foreground targets. This raises the question: Is there a method that can automatically aggregate similar objects within scenographic remote sensing images, thereby enabling models to differentiate knowledge embedded in various geospatial patterns for improved feature representation? In this work, we present the Pattern Integration and Enhancement Vision Transformer (PIEViT), a novel self-supervised learning framework designed specifically for remote sensing imagery. PIEViT utilizes a teacher-student architecture to address both image-level and patch-level tasks. It employs the Geospatial Pattern Cohesion (GPC) module to explore the natural clustering of patches, enhancing the differentiation of individual features. The Feature Integration Projection (FIP) module further refines masked token reconstruction using geospatially clustered patches. We validated PIEViT across multiple downstream tasks, including object detection, semantic segmentation, and change detection. Experiments demonstrated that PIEViT enhances the representation of internal patch features, providing significant improvements over existing self-supervised baselines. It achieves excellent results in object detection, land cover classification, and change detection, underscoring its robustness, generalization, and transferability for remote sensing image interpretation tasks.

cs.CV↗