Search

Large multimodal models still struggle with text-rich images because of inadequate training data. Self-Instruct provides an annotation-free way for generating instruction data, but its quality is poor, as multimodal alignment remains a hurdle even for the largest models. In this work, we propose LLaVAR-2, to enhance multimodal alignment for text-rich images through hybrid instruction generation between human annotators and large language models. Specifically, it involves detailed image captions from human annotators, followed by the use of these annotations in tailored text prompts for GPT-4o to curate a dataset. It also implements several mechanisms to filter out low-quality data, and the resulting dataset comprises 424k high-quality pairs of instructions. Empirical results show that models fine-tuned on this dataset exhibit impressive enhancements over those trained with self-instruct data., Comment: COLING 2025

Diffusion models have emerged as a powerful foundation model for visual generation. With an appropriate sampling process, it can effectively serve as a generative prior to solve general inverse problems. Current posterior sampling based methods take the measurement (i.e., degraded image sample) into the posterior sampling to infer the distribution of the target data (i.e., clean image sample). However, in this manner, we show that high-frequency information can be prematurely introduced during the early stages, which could induce larger posterior estimate errors during the restoration sampling. To address this issue, we first reveal that forming the log posterior gradient with the noisy measurement ( i.e., samples from a diffusion forward process) instead of the clean one can benefit the reverse process. Consequently, we propose a novel diffusion posterior sampling method DPS-CM, which incorporates a Crafted Measurement (i.e., samples generated by a reverse denoising process, compared to random sampling with noise in standard methods) to form the posterior estimate. This integration aims to mitigate the misalignment with the diffusion prior caused by cumulative posterior estimate errors. Experimental results demonstrate that our approach significantly improves the overall capacity to solve general and noisy inverse problems, such as Gaussian deblurring, super-resolution, inpainting, nonlinear deblurring, and tasks with Poisson noise, relative to existing approaches.

Reconstructing and understanding 3D structures from a limited number of images is a well-established problem in computer vision. Traditional methods usually break this task into multiple subtasks, each requiring complex transformations between different data representations. For instance, dense reconstruction through Structure-from-Motion (SfM) involves converting images into key points, optimizing camera parameters, and estimating structures. Afterward, accurate sparse reconstructions are required for further dense modeling, which is subsequently fed into task-specific neural networks. This multi-step process results in considerable processing time and increased engineering complexity. In this work, we present the Large Spatial Model (LSM), which processes unposed RGB images directly into semantic radiance fields. LSM simultaneously estimates geometry, appearance, and semantics in a single feed-forward operation, and it can generate versatile label maps by interacting with language at novel viewpoints. Leveraging a Transformer-based architecture, LSM integrates global geometry through pixel-aligned point maps. To enhance spatial attribute regression, we incorporate local context aggregation with multi-scale fusion, improving the accuracy of fine local details. To tackle the scarcity of labeled 3D semantic data and enable natural language-driven scene manipulation, we incorporate a pre-trained 2D language-based segmentation model into a 3D-consistent semantic feature field. An efficient decoder then parameterizes a set of semantic anisotropic Gaussians, facilitating supervised end-to-end learning. Extensive experiments across various tasks show that LSM unifies multiple 3D vision tasks directly from unposed images, achieving real-time semantic 3D reconstruction for the first time., Comment: Project Website: https://largespatialmodel.github.io

Wildfires have significantly increased in the United States (U.S.), making certain areas harder to live in. This motivates us to jointly analyze active fires and population changes in the U.S. from July 2020 to June 2021. The available data are recorded on different scales (or spatial resolutions) and by different types of distributions (referred to as multi-type data). Moreover, wildfires are known to have feedback mechanism that creates signal-to-noise dependence. We analyze point-referenced remote sensing fire data from National Aeronautics and Space Administration (NASA) and county-level population change data provided by U.S. Census Bureau's Population Estimates Program (PEP). We develop a multiscale multi-type spatial Bayesian model that assumes the average number of fires is zero-inflated normal, the incidence of fire as Bernoulli, and the percentage population change as normally distributed. This high-dimensional dataset makes Markov chain Monte Carlo (MCMC) implementation infeasible. We bypass MCMC by extending a recently introduced computationally efficient Bayesian framework to directly sample from the exact posterior distribution, which includes a term to model signal-to-noise dependence. Such signal-to-noise dependence is known to be present in wildfire data, but is commonly not accounted for. A simulation study is used to highlight the computational performance of our method. In our analysis, we obtained predictions of wildfire probabilities, identified several useful covariates, and found that regions with many fires were associated with population change.

The blooming of virtual reality and augmented reality (VR/AR) technologies has driven an increasing demand for the creation of high-quality, immersive, and dynamic environments. However, existing generative techniques either focus solely on dynamic objects or perform outpainting from a single perspective image, failing to meet the requirements of VR/AR applications that need free-viewpoint, 360$^{\circ}$ virtual views where users can move in all directions. In this work, we tackle the challenging task of elevating a single panorama to an immersive 4D experience. For the first time, we demonstrate the capability to generate omnidirectional dynamic scenes with 360$^{\circ}$ views at 4K (4096 $\times$ 2048) resolution, thereby providing an immersive user experience. Our method introduces a pipeline that facilitates natural scene animations and optimizes a set of dynamic Gaussians using efficient splatting techniques for real-time exploration. To overcome the lack of scene-scale annotated 4D data and models, especially in panoramic formats, we propose a novel \textbf{Panoramic Denoiser} that adapts generic 2D diffusion priors to animate consistently in 360$^{\circ}$ images, transforming them into panoramic videos with dynamic scenes at targeted regions. Subsequently, we propose \textbf{Dynamic Panoramic Lifting} to elevate the panoramic video into a 4D immersive environment while preserving spatial and temporal consistency. By transferring prior knowledge from 2D models in the perspective domain to the panoramic domain and the 4D lifting with spatial appearance and geometry regularization, we achieve high-quality Panorama-to-4D generation at a resolution of 4K for the first time.

Investigating the network stability or synchronization dynamics of multi-agent systems with time delays is of significant importance in numerous real-world applications. Such investigations often rely on solving the transcendental characteristic equations (TCEs) obtained from linearization of the considered systems around specific solutions. While stability results based on the TCEs with real-valued coefficients induced by symmetric networks in time-delayed models have been extensively explored in the literature, there remains a notable gap in stability analysis for the TCEs with complexvalued coefficients arising from asymmetric networked dynamics with time delays. To address this challenge comprehensively, we propose a rigorously geometric approach. By identifying and studying the stability crossing curves in the complex plane, we are able to determine the stability region of these systems. This approach is not only suitable for analyzing the stability of models with discrete time delays but also for models with various types of delays, including distributed time delays. Additionally, it can also handle random networks. We demonstrate the efficacy of this approach in designing delayed control strategies for car-following systems, mechanical systems, and deep brain stimulation modeling, where involved are complex-valued TCEs or/and different types of delays. All these therefore highlight the broad applicability of our approach across diverse domains., Comment: No

Enhancing our understanding of adversarial examples is crucial for the secure application of machine learning models in real-world scenarios. A prevalent method for analyzing adversarial examples is through a frequency-based approach. However, existing research indicates that attacks designed to exploit low-frequency or high-frequency information can enhance attack performance, leading to an unclear relationship between adversarial perturbations and different frequency components. In this paper, we seek to demystify this relationship by exploring the characteristics of adversarial perturbations within the frequency domain. We employ wavelet packet decomposition for detailed frequency analysis of adversarial examples and conduct statistical examinations across various frequency bands. Intriguingly, our findings indicate that significant adversarial perturbations are present within the high-frequency components of low-frequency bands. Drawing on this insight, we propose a black-box adversarial attack algorithm based on combining different frequency bands. Experiments conducted on multiple datasets and models demonstrate that combining low-frequency bands and high-frequency components of low-frequency bands can significantly enhance attack efficiency. The average attack success rate reaches 99\%, surpassing attacks that utilize a single frequency segment. Additionally, we introduce the normalized disturbance visibility index as a solution to the limitations of $L_2$ norm in assessing continuous and discrete perturbations.

The increasing demand for virtual reality applications has highlighted the significance of crafting immersive 3D assets. We present a text-to-3D 360$^{\circ}$ scene generation pipeline that facilitates the creation of comprehensive 360$^{\circ}$ scenes for in-the-wild environments in a matter of minutes. Our approach utilizes the generative power of a 2D diffusion model and prompt self-refinement to create a high-quality and globally coherent panoramic image. This image acts as a preliminary "flat" (2D) scene representation. Subsequently, it is lifted into 3D Gaussians, employing splatting techniques to enable real-time exploration. To produce consistent 3D geometry, our pipeline constructs a spatially coherent structure by aligning the 2D monocular depth into a globally optimized point cloud. This point cloud serves as the initial state for the centroids of 3D Gaussians. In order to address invisible issues inherent in single-view inputs, we impose semantic and geometric constraints on both synthesized and input camera views as regularizations. These guide the optimization of Gaussians, aiding in the reconstruction of unseen regions. In summary, our method offers a globally consistent 3D scene within a 360$^{\circ}$ perspective, providing an enhanced immersive experience over existing techniques. Project website at: http://dreamscene360.github.io/

Adversarial attacks in visual object tracking have significantly degraded the performance of advanced trackers by introducing imperceptible perturbations into images. However, there is still a lack of research on designing adversarial defense methods for object tracking. To address these issues, we propose an effective auxiliary pre-processing defense network, AADN, which performs defensive transformations on the input images before feeding them into the tracker. Moreover, it can be seamlessly integrated with other visual trackers as a plug-and-play module without parameter adjustments. We train AADN using adversarial training, specifically employing Dua-Loss to generate adversarial samples that simultaneously attack the classification and regression branches of the tracker. Extensive experiments conducted on the OTB100, LaSOT, and VOT2018 benchmarks demonstrate that AADN maintains excellent defense robustness against adversarial attack methods in both adaptive and non-adaptive attack scenarios. Moreover, when transferring the defense network to heterogeneous trackers, it exhibits reliable transferability. Finally, AADN achieves a processing time of up to 5ms/frame, allowing seamless integration with existing high-speed trackers without introducing significant computational overhead., Comment: This paper is accepted by ECCV2024

We introduce a novel machine unlearning framework founded upon the established principles of the min-max optimization paradigm. We capitalize on the capabilities of strong Membership Inference Attacks (MIA) to facilitate the unlearning of specific samples from a trained model. We consider the scenario of two networks, the attacker $\mathbf{A}$ and the trained defender $\mathbf{D}$ pitted against each other in an adversarial objective, wherein the attacker aims at teasing out the information of the data to be unlearned in order to infer membership, and the defender unlearns to defend the network against the attack, whilst preserving its general performance. The algorithm can be trained end-to-end using backpropagation, following the well known iterative min-max approach in updating the attacker and the defender. We additionally incorporate a self-supervised objective effectively addressing the feature space discrepancies between the forget set and the validation set, enhancing unlearning performance. Our proposed algorithm closely approximates the ideal benchmark of retraining from scratch for both random sample forgetting and class-wise forgetting schemes on standard machine-unlearning datasets. Specifically, on the class unlearning scheme, the method demonstrates near-optimal performance and comprehensively overcomes known methods over the random sample forgetting scheme across all metrics and multiple network pruning strategies., Comment: 13 pages including references, 2 tables, 2 figures and 1 algorithm

3D scene representations have gained immense popularity in recent years. Methods that use Neural Radiance fields are versatile for traditional tasks such as novel view synthesis. In recent times, some work has emerged that aims to extend the functionality of NeRF beyond view synthesis, for semantically aware tasks such as editing and segmentation using 3D feature field distillation from 2D foundation models. However, these methods have two major limitations: (a) they are limited by the rendering speed of NeRF pipelines, and (b) implicitly represented feature fields suffer from continuity artifacts reducing feature quality. Recently, 3D Gaussian Splatting has shown state-of-the-art performance on real-time radiance field rendering. In this work, we go one step further: in addition to radiance field rendering, we enable 3D Gaussian splatting on arbitrary-dimension semantic features via 2D foundation model distillation. This translation is not straightforward: naively incorporating feature fields in the 3DGS framework encounters significant challenges, notably the disparities in spatial resolution and channel consistency between RGB images and feature maps. We propose architectural and training changes to efficiently avert this problem. Our proposed method is general, and our experiments showcase novel view semantic segmentation, language-guided editing and segment anything through learning feature fields from state-of-the-art 2D foundation models such as SAM and CLIP-LSeg. Across experiments, our distillation method is able to provide comparable or better results, while being significantly faster to both train and render. Additionally, to the best of our knowledge, we are the first method to enable point and bounding-box prompting for radiance field manipulation, by leveraging the SAM model. Project website at: https://feature-3dgs.github.io/

In public health applications, spatial data collected are often recorded at different spatial scales and over different correlated variables. Spatial change of support is a key inferential problem in these applications and have become standard in univariate settings; however, it is less standard in multivariate settings. There are several existing multivariate spatial models that can be easily combined with multiscale spatial approach to analyze multivariate multiscale spatial data. In this paper, we propose three new models from such combinations for bivariate multiscale spatial data in a Bayesian context. In particular, we extend spatial random effects models, multivariate conditional autoregressive models, and ordered hierarchical models through a multiscale spatial approach. We run simulation studies for the three models and compare them in terms of prediction performance and computational efficiency. We motivate our models through an analysis of 2015 Texas annual average percentage receiving two blood tests from the Dartmouth Atlas Project.

The G-Brownian-motion-driven stochastic differential equations (G-SDEs) as well as the G-expectation, which were seminally proposed by Peng and his colleagues, have been extensively applied to describing a particular kind of uncertainty arising in real-world systems modeling. Mathematically depicting long-time and limit behaviors of the solution produced by G-SDEs is beneficial to understanding the mechanisms of system's evolution. Here, we develop a new G-semimartingale convergence theorem and further establish a new invariance principle for investigating the long-time behaviors emergent in G-SDEs. We also validate the uniqueness and the global existence of the solution of G-SDEs whose vector fields are only locally Lipchitzian with a linear upper bound. To demonstrate the broad applicability of our analytically established results, we investigate its application to achieving G-stochastic control in a few representative dynamical systems.

This paper is devoted to discuss the stabilizability of a class of $ 2 \times2 $ non-homogeneous hyperbolic systems. Motivated by the example in \cite[Page 197]{CB2016}, we analyze the influence of the interval length $L$ on stabilizability of the system. By spectral analysis, we prove that either the system is stabilizable for all $L>0$ or it possesses the dichotomy property: there exists a critical length $L_c>0$ such that the system is stabilizable for $L\in (0,L_c)$ but unstabilizable for $L\in [L_c,+\infty)$. In addition, for $L\in [L_c,+\infty)$, we obtain that the system can reach equilibrium state in finite time by backstepping control combined with observer. Finally, we also provide some numerical simulations to confirm our developed analytical criteria.

In our manuscript, we develop a new approach for stability analysis of one-dimensional wave equation with time delay. The major contribution of our work is to develop a new method for spectral analysis. We derive sufficient and necessary conditions for the feedback gain and time delay which guarantee the exponential stability of the closed-loop system. Comparing with similar conditions developed in the past literatures, we discuss all the situation when the time delay is positive, including when it is irrational. We prove that the exponential stability can be achieved if and only if the time delay is an even number. We also get the general formula term of the stability region of the coupling gain for different even multiples of time delay, and from this we easily obtain the shrink of the stability region as time delay increases. In addition, we explore the impact of slight perturbations in time delay on high frequency robustness.

Controlling complex dynamical systems has been a topic of considerable interest in academic circles in recent decades. While existing works have primarily focused on closed-loop control schemes with infinite-time durations, this paper introduces a novel finite-time, closed-loop stochastic controller that pays special attention to control time and energy and their dependence on system parameters. This technique of stochastic control not only enables finite-time control in chaotic dynamical systems but also facilitates finite-time synchronization in unidirectionally coupled systems. Notably, our new scheme offers several advantages over existing deterministic finite-time controllers from a physical standpoint of time and energy consumption. Using numerical experiments based on random ecosystems, neural networks, and Lorenz systems, we demonstrate the effectiveness of our analytical results. It is anticipated that this proposed stochastic scheme will have widespread applicability in controlling complex dynamical systems and achieving network synchronization., Comment: There are some mistakes in the paper. We need to revise it

This work introduces alternating latent topologies (ALTO) for high-fidelity reconstruction of implicit 3D surfaces from noisy point clouds. Previous work identifies that the spatial arrangement of latent encodings is important to recover detail. One school of thought is to encode a latent vector for each point (point latents). Another school of thought is to project point latents into a grid (grid latents) which could be a voxel grid or triplane grid. Each school of thought has tradeoffs. Grid latents are coarse and lose high-frequency detail. In contrast, point latents preserve detail. However, point latents are more difficult to decode into a surface, and quality and runtime suffer. In this paper, we propose ALTO to sequentially alternate between geometric representations, before converging to an easy-to-decode latent. We find that this preserves spatial expressiveness and makes decoding lightweight. We validate ALTO on implicit 3D recovery and observe not only a performance improvement over the state-of-the-art, but a runtime improvement of 3-10$\times$. Project website at https://visual.ee.ucla.edu/alto.htm/.

Link prediction is an important task that has wide applications in various domains. However, the majority of existing link prediction approaches assume the given graph follows homophily assumption, and designs similarity-based heuristics or representation learning approaches to predict links. However, many real-world graphs are heterophilic graphs, where the homophily assumption does not hold, which challenges existing link prediction methods. Generally, in heterophilic graphs, there are many latent factors causing the link formation, and two linked nodes tend to be similar in one or two factors but might be dissimilar in other factors, leading to low overall similarity. Thus, one way is to learn disentangled representation for each node with each vector capturing the latent representation of a node on one factor, which paves a way to model the link formation in heterophilic graphs, resulting in better node representation learning and link prediction performance. However, the work on this is rather limited. Therefore, in this paper, we study a novel problem of exploring disentangled representation learning for link prediction on heterophilic graphs. We propose a novel framework DisenLink which can learn disentangled representations by modeling the link formation and perform factor-aware message-passing to facilitate link prediction. Extensive experiments on 13 real-world datasets demonstrate the effectiveness of DisenLink for link prediction on both heterophilic and hemophiliac graphs. Our codes are available at https://github.com/sjz5202/DisenLink