arXiv subjects

Wei Suo

Publications and source records attributed to Wei Suo.

18 recordsLinked to original sources

arXiv2026

GSMap: 2D Gaussians for Online HD Mapping

Accurate High-Definition (HD) map construction is critical for autonomous driving, yet existing methods face a fundamental trade-off: vectorization-based approaches preserve topology but struggle with geometric fidelity, while rasterization-based approaches enable precise geometric supervision but produce unstructured outputs. To bridge this gap, we propose GSMap, a novel framework that unifies both paradigms via a learnable 2D Gaussian representation. Each map element is modeled as an ordered sequence of 2D Gaussians, whose centers correspond to the vertices of the vectorized polyline/polygon. This formulation enables simultaneous optimization through: (1) Differentiable rasterization that enforces pixel-level geometric constraints, and (2) Topology-aware vectorization that maintains structural regularity. Experiments on both nuScenes and Argoverse2 demonstrate that our Gaussian-based representation effectively unifies geometric and topological learning, achieving significant performance improvements and demonstrating strong compatibility with existing HD mapping architectures. Code will be available at https://github.com/peakpang/GSMap

cs.CV↗

arXiv2026

Optimization-Based Discovery of A Non-Attracting Flow State in An Oscillating-Cylinder Wake

In the flow past a stationary circular cylinder, the classical Karman vortex street arises from a Hopf bifurcation of the steady flow at the critical Reynolds number. Although this solution becomes dynamically unstable beyond this point, it remains an exact solution of the governing equations. Motivated by this observation, we investigates whether similar non-attracting flow solutions exist in the flow past a forced oscillating cylinder at supercritical Re. In the present study, while employing PINNs to investigate the flow past a forced oscillating cylinder, we identify a class of flow solutions that are inaccessible through direct time-stepping simulations. The obtained solution remains phase-locked with the cylinder oscillation frequency, despite the corresponding parameters lying outside the lock-in regime. To verify this solution, the obtained PINNs solution is used as the initial guess for an optimization based on the optimizing a discrete loss (ODIL) framework. The results show that the solution can be consistently maintained during the optimization process. This indicates that the solution is self-consistent in the optimization sense, although it does not an attracting state of the original dynamical system. To understand the reason, we compare the numerical evolution mechanisms of each solvers. The results indicate that, flow states that satisfy the governing equations but are dynamically non-attracting can be identified and maintained as minima of the optimization problem. For the flow past a forced oscillating cylinder, non-attracting periodic solutions that satisfy the governing equations exist in addition to the attracting states obtained by conventional time-stepping simulations. Optimization-based solvers can therefore reveal such flow states that are difficult to obtain through direct time integration, providing a new perspective for understanding complex wake dynamics.

physics.flu-dyn↗

arXiv2026

Hallucination-aware intermediate representation edit in large vision-language models

Large Vision-Language Models have demonstrated exceptional performance in multimodal reasoning and complex scene understanding. However, these models still face significant hallucination issues, where outputs contradict visual facts. Recent research on hallucination mitigation has focused on retraining methods and Contrastive Decoding (CD) methods. While both methods perform well, retraining methods require substantial training resources, and CD methods introduce dual inference overhead. These factors hinder their practical applicability. To address the above issue, we propose a framework for dynamically detecting hallucination representations and performing hallucination-eliminating edits on these representations. With minimal additional computational cost, we achieve state-of-the-art performance on existing benchmarks. Extensive experiments demonstrate the effectiveness of our approach, highlighting its efficient and robust hallucination elimination capability and its powerful controllability over hallucinations. Code is available at https://github.com/ASGO-MM/HIRE

cs.CV↗

arXiv2026

Understanding and Mitigating Hallucinations in Multimodal Chain-of-Thought Models

Multimodal Chain-of-Thought (MCoT) models have demonstrated impressive capability in complex visual reasoning tasks. Unfortunately, recent studies reveal that they suffer from severe hallucination problems due to diminished visual attention during the generation process. However, visual attention decay is a well-studied problem in Large Vision-Language Models (LVLMs). Considering the fundamental differences in reasoning processes between MCoT models and traditional LVLMs, we raise a basic question: Whether MCoT models have unique causes of hallucinations? To answer this question, we systematically investigate the hallucination patterns of MCoT models and find that fabricated texts are primarily generated in associative reasoning steps, which we term divergent thinking. Leveraging these insights, we introduce a simple yet effective strategy that can effectively localize divergent thinking steps and intervene in the decoding process to mitigate hallucinations. Extensive experiments show that our method outperforms existing methods by a large margin. More importantly, our proposed method can be conveniently integrated with other hallucination mitigation methods and further boost their performance. The code is publicly available at https://github.com/ASGO-MM/MCoT-hallucination.

cs.CV↗

arXiv2025

Mitigating Information Loss under High Pruning Rates for Efficient Large Vision Language Models

Despite the great success of Large Vision Language Models (LVLMs), their high computational cost severely limits their broad applications. The computational cost of LVLMs mainly stems from the visual sequence of the input, which consists of hundreds or even thousands of tokens. Although existing methods have made progress by removing redundant tokens, they suffer from severe performance degradation with high pruning rates due to the loss of visual information. In this paper, we propose an Adaptive Content Compensation Method (ACCM), which can effectively mitigate the visual information loss via an image caption. Specifically, ACCM comprises two key components: a lightweight caption model and a selector. Firstly the caption model generates question-related descriptions under the guidance of the user instruction. Then the selector further identifies a contextually appropriate caption from multiple candidates. Leveraging self-supervised learning, our modules could be learned efficiently without any human or automated labeling. We conduct extensive experiments across seven benchmarks and the results show that ACCM significantly outperforms existing methods with lower FLOPs (e.g., surpassing SOTA by 20.6% with 6.5% fewer FLOPs).

cs.CV↗

arXiv2025

Short-LVLM: Compressing and Accelerating Large Vision-Language Models by Pruning Redundant Layers

Although large vision-language models (LVLMs) have demonstrated impressive capabilities in multi-modal understanding and reasoning, their practical applications are still limited by massive model parameters and high computational costs. Recent efforts from natural language processing (NLP) have shown the effectiveness of layer pruning, offering a plausible training-free compression solution. However, due to the modality divergence between vision and language, it is unclear whether these NLP techniques are still effective in LVLMs. In this paper, we empirically prove that directly applying these layer pruning methods to LVLMs is ineffective. Through extensive experiments, we find that non-essential vision-language (VL) tokens and inter-layer feature gaps pose critical challenges to pruning layers in LVLMs. Based on these insights, we propose a novel framework Short-LVLM (SVL) that can utilize important VL tokens and mitigate the layer-wise feature gaps. Notably, Short-LVLM not only achieves a superior trade-off between performance and efficiency but also exhibits several potential advantages, i.e., training-free, model-agnostic, and highly compatible. The code for this work is publicly available at https://github.com/ASGO-MM/Short-LVLM.

cs.CV↗

arXiv2025

Octopus: Alleviating Hallucination via Dynamic Contrastive Decoding

Large Vision-Language Models (LVLMs) have obtained impressive performance in visual content understanding and multi-modal reasoning. Unfortunately, these large models suffer from serious hallucination problems and tend to generate fabricated responses. Recently, several Contrastive Decoding (CD) strategies have been proposed to alleviate hallucination by introducing disturbed inputs. Although great progress has been made, these CD strategies mostly apply a one-size-fits-all approach for all input conditions. In this paper, we revisit this process through extensive experiments. Related results show that hallucination causes are hybrid and each generative step faces a unique hallucination challenge. Leveraging these meaningful insights, we introduce a simple yet effective Octopus-like framework that enables the model to adaptively identify hallucination types and create a dynamic CD workflow. Our Octopus framework not only outperforms existing methods across four benchmarks but also demonstrates excellent deployability and expansibility. Code is available at https://github.com/LijunZhang01/Octopus.

cs.CV↗

arXiv2024

Pruning All-Rounder: Rethinking and Improving Inference Efficiency for Large Vision Language Models

Although Large Vision-Language Models (LVLMs) have achieved impressive results, their high computational costs pose a significant barrier to wide application. To enhance inference efficiency, most existing approaches can be categorized as parameter-dependent or token-dependent strategies to reduce computational demands. However, parameter-dependent methods require retraining LVLMs to recover performance while token-dependent strategies struggle to consistently select the most relevant tokens. In this paper, we systematically analyze the above challenges and provide a series of valuable insights for inference acceleration. Based on these findings, we propose a novel framework, the Pruning All-Rounder (PAR). Different from previous works, PAR develops a meta-router to adaptively organize pruning flows across both tokens and layers. With a self-supervised learning manner, our method achieves a superior balance between performance and efficiency. Notably, PAR is highly flexible, offering multiple pruning versions to address a range of acceleration scenarios. The code for this work is publicly available at https://github.com/ASGO-MM/Pruning-All-Rounder.

cs.CV↗

arXiv2024

Physics Informed Neural Networks (PINNs) as intelligent computing technique for solving partial differential equations: Limitation and Future prospects

In recent years, Physics-Informed Neural Networks (PINNs) have become a representative method for solving partial differential equations (PDEs) with neural networks. PINNs provide a novel approach to solving PDEs through optimization algorithms, offering a unified framework for solving both forward and inverse problems. However, some limitations in terms of solution accuracy and generality have also been revealed. This paper systematically summarizes the limitations of PINNs and identifies three root causes for their failure in solving PDEs: (1) Poor multiscale approximation ability and ill-conditioning caused by PDE losses; (2) Insufficient exploration of convergence and error analysis, resulting in weak mathematical rigor; (3) Inadequate integration of physical information, causing mismatch between residuals and iteration errors. By focusing on addressing these limitations in PINNs, we outline the future directions and prospects for the intelligent computing of PDEs: (1) Analysis of ill-conditioning in PINNs and mitigation strategies; (2) Improvements to PINNs by enforcing temporal causality; (3) Empowering PINNs with classical numerical methods.

physics.comp-ph↗

arXiv2024

PINN-MG: A Multigrid-Inspired Hybrid Framework Combining Iterative Method and Physics-Informed Neural Networks

Iterative methods are widely used for solving partial differential equations (PDEs). However, the difficulty in eliminating global low-frequency errors significantly limits their convergence speed. In recent years, neural networks have emerged as a novel approach for solving PDEs, with studies revealing that they exhibit faster convergence for low-frequency components. Building on this complementary frequency convergence characteristics of iterative methods and neural networks, we draw inspiration from multigrid methods and propose a hybrid solving framework that combining iterative methods and neural network-based solvers, termed PINN-MG (PMG). In this framework, the iterative method is responsible for eliminating local high-frequency oscillation errors, while Physics-Informed Neural Networks (PINNs) are employed to correct global low-frequency errors. Throughout the solving process, high- and low-frequency components alternately dominate the error, with each being addressed by the iterative method and PINNs respectively, thereby accelerating the convergence. We tested the proposed PMG framework on the linear Poisson equation and the nonlinear Helmholtz equation, and the results demonstrated significant acceleration of the PMG when built on Gauss-Seidel, pseudo-time, and GMRES methods. Furthermore, detailed analysis of the convergence process further validates the rationality of the framework. We proposed that the PMG framework is a hybrid solving approach that does not rely on training data, achieving an organic integration of neural network methods with iterative methods.

physics.comp-ph↗

arXiv2024

A Plug-and-Play Method for Rare Human-Object Interactions Detection by Bridging Domain Gap

Human-object interactions (HOI) detection aims at capturing human-object pairs in images and corresponding actions. It is an important step toward high-level visual reasoning and scene understanding. However, due to the natural bias from the real world, existing methods mostly struggle with rare human-object pairs and lead to sub-optimal results. Recently, with the development of the generative model, a straightforward approach is to construct a more balanced dataset based on a group of supplementary samples. Unfortunately, there is a significant domain gap between the generated data and the original data, and simply merging the generated images into the original dataset cannot significantly boost the performance. To alleviate the above problem, we present a novel model-agnostic framework called \textbf{C}ontext-\textbf{E}nhanced \textbf{F}eature \textbf{A}lignment (CEFA) module, which can effectively align the generated data with the original data at the feature level and bridge the domain gap. Specifically, CEFA consists of a feature alignment module and a context enhancement module. On one hand, considering the crucial role of human-object pairs information in HOI tasks, the feature alignment module aligns the human-object pairs by aggregating instance information. On the other hand, to mitigate the issue of losing important context information caused by the traditional discriminator-style alignment method, we employ a context-enhanced image reconstruction module to improve the model's learning ability of contextual cues. Extensive experiments have shown that our method can serve as a plug-and-play module to improve the detection performance of HOI models on rare categories\footnote{https://github.com/LijunZhang01/CEFA}.

cs.CV↗

arXiv2024

Visual Prompt Selection for In-Context Learning Segmentation

As a fundamental and extensively studied task in computer vision, image segmentation aims to locate and identify different semantic concepts at the pixel level. Recently, inspired by In-Context Learning (ICL), several generalist segmentation frameworks have been proposed, providing a promising paradigm for segmenting specific objects. However, existing works mostly ignore the value of visual prompts or simply apply similarity sorting to select contextual examples. In this paper, we focus on rethinking and improving the example selection strategy. By comprehensive comparisons, we first demonstrate that ICL-based segmentation models are sensitive to different contexts. Furthermore, empirical evidence indicates that the diversity of contextual prompts plays a crucial role in guiding segmentation. Based on the above insights, we propose a new stepwise context search method. Different from previous works, we construct a small yet rich candidate pool and adaptively search the well-matched contexts. More importantly, this method effectively reduces the annotation cost by compacting the search space. Extensive experiments show that our method is an effective strategy for selecting examples and enhancing segmentation performance.

cs.CV↗

arXiv2024

C3L: Content Correlated Vision-Language Instruction Tuning Data Generation via Contrastive Learning

Vision-Language Instruction Tuning (VLIT) is a critical training phase for Large Vision-Language Models (LVLMs). With the improving capabilities of open-source LVLMs, researchers have increasingly turned to generate VLIT data by using open-source LVLMs and achieved significant progress. However, such data generation approaches are bottlenecked by the following challenges: 1) Since multi-modal models tend to be influenced by prior language knowledge, directly using LVLMs to generate VLIT data would inevitably lead to low content relevance between generated data and images. 2) To improve the ability of the models to generate VLIT data, previous methods have incorporated an additional training phase to boost the generative capacity. This process hurts the generalization of the models to unseen inputs (i.e., "exposure bias" problem). In this paper, we propose a new Content Correlated VLIT data generation via Contrastive Learning (C3L). Specifically, we design a new content relevance module which enhances the content relevance between VLIT data and images by computing Image Instruction Correspondence Scores S(I2C). Moreover, a contrastive learning module is introduced to further boost the VLIT data generation capability of the LVLMs. A large number of automatic measures on four benchmarks show the effectiveness of our method.

cs.CV↗

arXiv2024

CoLeCLIP: Open-Domain Continual Learning via Joint Task Prompt and Vocabulary Learning

This paper explores the problem of continual learning (CL) of vision-language models (VLMs) in open domains, where the models need to perform continual updating and inference on a streaming of datasets from diverse seen and unseen domains with novel classes. Such a capability is crucial for various applications in open environments, e.g., AI assistants, autonomous driving systems, and robotics. Current CL studies mostly focus on closed-set scenarios in a single domain with known classes. Large pre-trained VLMs like CLIP have demonstrated superior zero-shot recognition ability, and a number of recent studies leverage this ability to mitigate catastrophic forgetting in CL, but they focus on closed-set CL in a single domain dataset. Open-domain CL of large VLMs is significantly more challenging due to 1) large class correlations and domain gaps across the datasets and 2) the forgetting of zero-shot knowledge in the pre-trained VLMs in addition to the knowledge learned from the newly adapted datasets. In this work we introduce a novel approach, termed CoLeCLIP, that learns an open-domain CL model based on CLIP. It addresses these challenges by a joint learning of a set of task prompts and a cross-domain class vocabulary. Extensive experiments on 11 domain datasets show that CoLeCLIP outperforms state-of-the-art methods for open-domain CL under both task- and class-incremental learning settings.

cs.CV↗

arXiv2023

A novel paradigm for solving PDEs: multi scale neural computing

Numerical simulation is dominant in solving partial difference equations (PDEs), but balancing fine-grained grids with low computational costs is challenging. Recently, solving PDEs with neural networks (NNs) has gained interest, yet cost-effectiveness and high accuracy remains a challenge. This work introduces a novel paradigm for solving PDEs, called multi scale neural computing (MSNC), considering spectral bias of NNs and local approximation properties in the finite difference method (FDM). The MSNC decomposes the solution with a NN for efficient capture of global scale and the FDM for detailed description of local scale, aiming to balance costs and accuracy. Demonstrated advantages include higher accuracy (10 times for 1D PDEs, 20 times for 2D PDEs) and lower costs (4 times for 1D PDEs, 16 times for 2D PDEs) than the standard FDM. The MSNC also exhibits stable convergence and rigorous boundary condition satisfaction, showcasing the potential for hybrid of NN and numerical method.

physics.comp-ph↗

arXiv2023

S3C: Semi-Supervised VQA Natural Language Explanation via Self-Critical Learning

VQA Natural Language Explanation (VQA-NLE) task aims to explain the decision-making process of VQA models in natural language. Unlike traditional attention or gradient analysis, free-text rationales can be easier to understand and gain users' trust. Existing methods mostly use post-hoc or self-rationalization models to obtain a plausible explanation. However, these frameworks are bottlenecked by the following challenges: 1) the reasoning process cannot be faithfully responded to and suffer from the problem of logical inconsistency. 2) Human-annotated explanations are expensive and time-consuming to collect. In this paper, we propose a new Semi-Supervised VQA-NLE via Self-Critical Learning (S3C), which evaluates the candidate explanations by answering rewards to improve the logical consistency between answers and rationales. With a semi-supervised learning framework, the S3C can benefit from a tremendous amount of samples without human-annotated explanations. A large number of automatic measures and human evaluations all show the effectiveness of our method. Meanwhile, the framework achieves a new state-of-the-art performance on the two VQA-NLE datasets.

cs.CV↗

arXiv2022

Dual-Level Decoupled Transformer for Video Captioning

Video captioning aims to understand the spatio-temporal semantic concept of the video and generate descriptive sentences. The de-facto approach to this task dictates a text generator to learn from \textit{offline-extracted} motion or appearance features from \textit{pre-trained} vision models. However, these methods may suffer from the so-called \textbf{\textit{"couple"}} drawbacks on both \textit{video spatio-temporal representation} and \textit{sentence generation}. For the former, \textbf{\textit{"couple"}} means learning spatio-temporal representation in a single model(3DCNN), resulting the problems named \emph{disconnection in task/pre-train domain} and \emph{hard for end-to-end training}. As for the latter, \textbf{\textit{"couple"}} means treating the generation of visual semantic and syntax-related words equally. To this end, we present $\mathcal{D}^{2}$ - a dual-level decoupled transformer pipeline to solve the above drawbacks: \emph{(i)} for video spatio-temporal representation, we decouple the process of it into "first-spatial-then-temporal" paradigm, releasing the potential of using dedicated model(\textit{e.g.} image-text pre-training) to connect the pre-training and downstream tasks, and makes the entire model end-to-end trainable. \emph{(ii)} for sentence generation, we propose \emph{Syntax-Aware Decoder} to dynamically measure the contribution of visual semantic and syntax-related words. Extensive experiments on three widely-used benchmarks (MSVD, MSR-VTT and VATEX) have shown great potential of the proposed $\mathcal{D}^{2}$ and surpassed the previous methods by a large margin in the task of video captioning.

cs.CV↗

arXiv2021

Proposal-free One-stage Referring Expression via Grid-Word Cross-Attention

Referring Expression Comprehension (REC) has become one of the most important tasks in visual reasoning, since it is an essential step for many vision-and-language tasks such as visual question answering. However, it has not been widely used in many downstream tasks because it suffers 1) two-stage methods exist heavy computation cost and inevitable error accumulation, and 2) one-stage methods have to depend on lots of hyper-parameters (such as anchors) to generate bounding box. In this paper, we present a proposal-free one-stage (PFOS) model that is able to regress the region-of-interest from the image, based on a textual query, in an end-to-end manner. Instead of using the dominant anchor proposal fashion, we directly take the dense-grid of an image as input for a cross-attention transformer that learns grid-word correspondences. The final bounding box is predicted directly from the image without the time-consuming anchor selection process that previous methods suffer. Our model achieves the state-of-the-art performance on four referring expression datasets with higher efficiency, comparing to previous best one-stage and two-stage methods.

cs.CV↗