id
stringlengths
9
16
title
stringlengths
4
278
categories
stringlengths
5
104
abstract
stringlengths
6
4.09k
2501.12194
An End-to-End Approach for Korean Wakeword Systems with Speaker Authentication
cs.SD cs.AI cs.LG eess.AS
Wakeword detection plays a critical role in enabling AI assistants to listen to user voices and interact effectively. However, for languages other than English, there is a significant lack of pre-trained wakeword models. Additionally, systems that merely determine the presence of a wakeword can pose serious privacy concerns. In this paper, we propose an end-to-end approach that trains wakewords for Non-English languages, particulary Korean, and uses this to develop a Voice Authentication model to protect user privacy. Our implementation employs an open-source platform OpenWakeWord, which performs wakeword detection using an FCN (Fully-Connected Network) architecture. Once a wakeword is detected, our custom-developed code calculates cosine similarity for robust user authentication. Experimental results demonstrate the effectiveness of our approach, achieving a 16.79% and a 6.6% Equal Error Rate (EER) each in the Wakeword Detection and the Voice Authentication. These findings highlight the model's potential in providing secure and accurate wakeword detection and authentication for Korean users.
2501.12198
Opinion dynamics in bounded confidence models with manipulative agents: Moving the Overton window
cs.SI physics.soc-ph stat.AP
This paper focuses on the opinion dynamics under the influence of manipulative agents. This type of agents is characterized by the fact that their opinions follow a trajectory that does not respond to the dynamics of the model, although it does influence the rest of the normal agents. Simulation has been implemented to study how one manipulative group modifies the natural dynamics of some opinion models of bounded confidence. It is studied what strategies based on the number of manipulative agents and their common opinion trajectory can be carried out by a manipulative group to influence normal agents and attract them to their opinions. In certain weighted models, some effects are observed in which normal agents move in the opposite direction to the manipulator group. Moreover, the conditions which ensure the influence of a manipulative group on a group of normal agents over time are also established for the Hegselmann-Krause model.
2501.12199
Experience-replay Innovative Dynamics
cs.LG cs.GT cs.MA
Despite its groundbreaking success, multi-agent reinforcement learning (MARL) still suffers from instability and nonstationarity. Replicator dynamics, the most well-known model from evolutionary game theory (EGT), provide a theoretical framework for the convergence of the trajectories to Nash equilibria and, as a result, have been used to ensure formal guarantees for MARL algorithms in stable game settings. However, they exhibit the opposite behavior in other settings, which poses the problem of finding alternatives to ensure convergence. In contrast, innovative dynamics, such as the Brown-von Neumann-Nash (BNN) or Smith, result in periodic trajectories with the potential to approximate Nash equilibria. Yet, no MARL algorithms based on these dynamics have been proposed. In response to this challenge, we develop a novel experience replay-based MARL algorithm that incorporates revision protocols as tunable hyperparameters. We demonstrate, by appropriately adjusting the revision protocols, that the behavior of our algorithm mirrors the trajectories resulting from these dynamics. Importantly, our contribution provides a framework capable of extending the theoretical guarantees of MARL algorithms beyond replicator dynamics. Finally, we corroborate our theoretical findings with empirical results.
2501.12202
Hunyuan3D 2.0: Scaling Diffusion Models for High Resolution Textured 3D Assets Generation
cs.CV
We present Hunyuan3D 2.0, an advanced large-scale 3D synthesis system for generating high-resolution textured 3D assets. This system includes two foundation components: a large-scale shape generation model -- Hunyuan3D-DiT, and a large-scale texture synthesis model -- Hunyuan3D-Paint. The shape generative model, built on a scalable flow-based diffusion transformer, aims to create geometry that properly aligns with a given condition image, laying a solid foundation for downstream applications. The texture synthesis model, benefiting from strong geometric and diffusion priors, produces high-resolution and vibrant texture maps for either generated or hand-crafted meshes. Furthermore, we build Hunyuan3D-Studio -- a versatile, user-friendly production platform that simplifies the re-creation process of 3D assets. It allows both professional and amateur users to manipulate or even animate their meshes efficiently. We systematically evaluate our models, showing that Hunyuan3D 2.0 outperforms previous state-of-the-art models, including the open-source models and closed-source models in geometry details, condition alignment, texture quality, and etc. Hunyuan3D 2.0 is publicly released in order to fill the gaps in the open-source 3D community for large-scale foundation generative models. The code and pre-trained weights of our models are available at: https://github.com/Tencent/Hunyuan3D-2
2501.12203
Explainability for Vision Foundation Models: A Survey
cs.CV
As artificial intelligence systems become increasingly integrated into daily life, the field of explainability has gained significant attention. This trend is particularly driven by the complexity of modern AI models and their decision-making processes. The advent of foundation models, characterized by their extensive generalization capabilities and emergent uses, has further complicated this landscape. Foundation models occupy an ambiguous position in the explainability domain: their complexity makes them inherently challenging to interpret, yet they are increasingly leveraged as tools to construct explainable models. In this survey, we explore the intersection of foundation models and eXplainable AI (XAI) in the vision domain. We begin by compiling a comprehensive corpus of papers that bridge these fields. Next, we categorize these works based on their architectural characteristics. We then discuss the challenges faced by current research in integrating XAI within foundation models. Furthermore, we review common evaluation methodologies for these combined approaches. Finally, we present key observations and insights from our survey, offering directions for future research in this rapidly evolving field.
2501.12204
Score Combining for Contrastive OOD Detection
cs.LG
In out-of-distribution (OOD) detection, one is asked to classify whether a test sample comes from a known inlier distribution or not. We focus on the case where the inlier distribution is defined by a training dataset and there exists no additional knowledge about the novelties that one is likely to encounter. This problem is also referred to as novelty detection, one-class classification, and unsupervised anomaly detection. The current literature suggests that contrastive learning techniques are state-of-the-art for OOD detection. We aim to improve on those techniques by combining/ensembling their scores using the framework of null hypothesis testing and, in particular, a novel generalized likelihood ratio test (GLRT). We demonstrate that our proposed GLRT-based technique outperforms the state-of-the-art CSI and SupCSI techniques from Tack et al. 2020 in dataset-vs-dataset experiments with CIFAR-10, SVHN, LSUN, ImageNet, and CIFAR-100, as well as leave-one-class-out experiments with CIFAR-10. We also demonstrate that our GLRT outperforms the score-combining methods of Fisher, Bonferroni, Simes, Benjamini-Hochwald, and Stouffer in our application.
2501.12206
Fixing Imbalanced Attention to Mitigate In-Context Hallucination of Large Vision-Language Model
cs.CV cs.CL
Large Vision Language Models (LVLMs) have demonstrated remarkable capabilities in understanding and describing visual content, achieving state-of-the-art performance across various vision-language tasks. However, these models frequently exhibit hallucination behavior, where they generate descriptions containing objects or details absent in the input image. Our work investigates this phenomenon by analyzing attention patterns across transformer layers and heads, revealing that hallucinations often stem from progressive degradation of visual grounding in deeper layers. We propose a novel attention modification approach that combines selective token emphasis and head-specific modulation to maintain visual grounding throughout the generation process. Our method introduces two key components: (1) a dual-stream token selection mechanism that identifies and prioritizes both locally informative and spatially significant visual tokens, and (2) an attention head-specific modulation strategy that differentially amplifies visual information processing based on measured visual sensitivity of individual attention heads. Through extensive experimentation on the MSCOCO dataset, we demonstrate that our approach reduces hallucination rates by up to 62.3\% compared to baseline models while maintaining comparable task performance. Our analysis reveals that selectively modulating tokens across attention heads with varying levels of visual sensitivity can significantly improve visual grounding without requiring model retraining.
2501.12208
Community Discovery Algorithm Based on Spatio-temporal Graph Embedding in Dynamic Social Networks
cs.SI
Community discovery is one of the key issues in the study of dynamic social networks. Traditional community discovery algorithms only focus on the establishment and disconnection of connections between nodes, failing to capture deeper factors. To address this limitation, in this work, we propose a community discovery algorithm based on spatiotemporal graph embedding (CDA-SGE), which integrates spatial information and evolutions of nodes to comprehensively capture the dynamic features of networks. Specifically, this algorithm employs Graph Convolutional Neural Networks (GCN) to aggregate latent spatial information, effectively representing the embedding of nodes in space. Temporal evolutions of the nodes are then modeled using Gated Recurrent Units (GRU), thereby solving problems such as node dynamism and relationship transmission. Finally, a Self-Organizing Map (SOM) is applied to cluster dynamic network representations and identify community affiliations of nodes. We then perform simulations on four types of dynamic networks and show that the CDA-SGE outperforms traditional community discovery algorithms in terms of purity, standardized mutual information, heterogeneity, and homogeneity. These results demonstrate the algorithm's superior ability to accurately uncover community structures hidden in dynamic social networks.
2501.12212
Quantitative Error Bounds for Scaling Limits of Stochastic Iterative Algorithms
stat.ML cs.LG math.PR math.ST stat.TH
Stochastic iterative algorithms, including stochastic gradient descent (SGD) and stochastic gradient Langevin dynamics (SGLD), are widely utilized for optimization and sampling in large-scale and high-dimensional problems in machine learning, statistics, and engineering. Numerous works have bounded the parameter error in, and characterized the uncertainty of, these approximations. One common approach has been to use scaling limit analyses to relate the distribution of algorithm sample paths to a continuous-time stochastic process approximation, particularly in asymptotic setups. Focusing on the univariate setting, in this paper, we build on previous work to derive non-asymptotic functional approximation error bounds between the algorithm sample paths and the Ornstein-Uhlenbeck approximation using an infinite-dimensional version of Stein's method of exchangeable pairs. We show that this bound implies weak convergence under modest additional assumptions and leads to a bound on the error of the variance of the iterate averages of the algorithm. Furthermore, we use our main result to construct error bounds in terms of two common metrics: the L\'{e}vy-Prokhorov and bounded Wasserstein distances. Our results provide a foundation for developing similar error bounds for the multivariate setting and for more sophisticated stochastic approximation algorithms.
2501.12214
Improving robot understanding using conversational AI: demonstration and feasibility study
cs.RO cs.HC
Explanations constitute an important aspect of successful human robot interactions and can enhance robot understanding. To improve the understanding of the robot, we have developed four levels of explanation (LOE) based on two questions: what needs to be explained, and why the robot has made a particular decision. The understandable robot requires a communicative action when there is disparity between the human s mental model of the robot and the robots state of mind. This communicative action was generated by utilizing a conversational AI platform to generate explanations. An adaptive dialog was implemented for transition from one LOE to another. Here, we demonstrate the adaptive dialog in a collaborative task with errors and provide results of a feasibility study with users.
2501.12215
Automatic selection of the best neural architecture for time series forecasting via multi-objective optimization and Pareto optimality conditions
cs.LG
Time series forecasting plays a pivotal role in a wide range of applications, including weather prediction, healthcare, structural health monitoring, predictive maintenance, energy systems, and financial markets. While models such as LSTM, GRU, Transformers, and State-Space Models (SSMs) have become standard tools in this domain, selecting the optimal architecture remains a challenge. Performance comparisons often depend on evaluation metrics and the datasets under analysis, making the choice of a universally optimal model controversial. In this work, we introduce a flexible automated framework for time series forecasting that systematically designs and evaluates diverse network architectures by integrating LSTM, GRU, multi-head Attention, and SSM blocks. Using a multi-objective optimization approach, our framework determines the number, sequence, and combination of blocks to align with specific requirements and evaluation objectives. From the resulting Pareto-optimal architectures, the best model for a given context is selected via a user-defined preference function. We validate our framework across four distinct real-world applications. Results show that a single-layer GRU or LSTM is usually optimal when minimizing training time alone. However, when maximizing accuracy or balancing multiple objectives, the best architectures are often composite designs incorporating multiple block types in specific configurations. By employing a weighted preference function, users can resolve trade-offs between objectives, revealing novel, context-specific optimal architectures. Our findings underscore that no single neural architecture is universally optimal for time series forecasting. Instead, the best-performing model emerges as a data-driven composite architecture tailored to user-defined criteria and evaluation objectives.
2501.12216
RL-RC-DoT: A Block-level RL agent for Task-Aware Video Compression
cs.LG cs.CV eess.IV
Video encoders optimize compression for human perception by minimizing reconstruction error under bit-rate constraints. In many modern applications such as autonomous driving, an overwhelming majority of videos serve as input for AI systems performing tasks like object recognition or segmentation, rather than being watched by humans. It is therefore useful to optimize the encoder for a downstream task instead of for perceptual image quality. However, a major challenge is how to combine such downstream optimization with existing standard video encoders, which are highly efficient and popular. Here, we address this challenge by controlling the Quantization Parameters (QPs) at the macro-block level to optimize the downstream task. This granular control allows us to prioritize encoding for task-relevant regions within each frame. We formulate this optimization problem as a Reinforcement Learning (RL) task, where the agent learns to balance long-term implications of choosing QPs on both task performance and bit-rate constraints. Notably, our policy does not require the downstream task as an input during inference, making it suitable for streaming applications and edge devices such as vehicles. We demonstrate significant improvements in two tasks, car detection, and ROI (saliency) encoding. Our approach improves task performance for a given bit rate compared to traditional task agnostic encoding methods, paving the way for more efficient task-aware video compression.
2501.12217
Early Detection and Classification of Breast Cancer Using Deep Learning Techniques
cs.CV cs.LG
Breast cancer is one of the deadliest cancers causing about massive number of patients to die annually all over the world according to the WHO. It is a kind of cancer that develops when the tissues of the breast grow rapidly and unboundly. This fatality rate can be prevented if the cancer is detected before it gets malignant. Using automation for early-age detection of breast cancer, Artificial Intelligence and Machine Learning technologies can be implemented for the best outcome. In this study, we are using the Breast Cancer Image Classification dataset collected from the Kaggle depository, which comprises 9248 Breast Ultrasound Images and is classified into three categories: Benign, Malignant, and Normal which refers to non-cancerous, cancerous, and normal images.This research introduces three pretrained model featuring custom classifiers that includes ResNet50, MobileNet, and VGG16, along with a custom CNN model utilizing the ReLU activation function.The models ResNet50, MobileNet, VGG16, and a custom CNN recorded accuracies of 98.41%, 97.91%, 98.19%, and 92.94% on the dataset, correspondingly, with ResNet50 achieving the highest accuracy of 98.41%.This model, with its deep and powerful architecture, is particularly successful in detecting aberrant cells as well as cancerous or non-cancerous tumors. These accuracies show that the Machine Learning methods are more compatible for the classification and early detection of breast cancer.
2501.12218
Exploring Temporally-Aware Features for Point Tracking
cs.CV
Point tracking in videos is a fundamental task with applications in robotics, video editing, and more. While many vision tasks benefit from pre-trained feature backbones to improve generalizability, point tracking has primarily relied on simpler backbones trained from scratch on synthetic data, which may limit robustness in real-world scenarios. Additionally, point tracking requires temporal awareness to ensure coherence across frames, but using temporally-aware features is still underexplored. Most current methods often employ a two-stage process: an initial coarse prediction followed by a refinement stage to inject temporal information and correct errors from the coarse stage. These approach, however, is computationally expensive and potentially redundant if the feature backbone itself captures sufficient temporal information. In this work, we introduce Chrono, a feature backbone specifically designed for point tracking with built-in temporal awareness. Leveraging pre-trained representations from self-supervised learner DINOv2 and enhanced with a temporal adapter, Chrono effectively captures long-term temporal context, enabling precise prediction even without the refinement stage. Experimental results demonstrate that Chrono achieves state-of-the-art performance in a refiner-free setting on the TAP-Vid-DAVIS and TAP-Vid-Kinetics datasets, among common feature backbones used in point tracking as well as DINOv2, with exceptional efficiency. Project page: https://cvlab-kaist.github.io/Chrono/
2501.12222
Strong phonon-mediated high temperature superconductivity in Li$_2$AuH$_6$ under ambient pressure
cond-mat.supr-con cond-mat.mtrl-sci cs.AI physics.comp-ph
We used our developed AI search engine~(InvDesFlow) to perform extensive investigations regarding ambient stable superconducting hydrides. A cubic structure Li$_2$AuH$_6$ with Au-H octahedral motifs is identified to be a candidate. After performing thermodynamical analysis, we provide a feasible route to experimentally synthesize this material via the known LiAu and LiH compounds under ambient pressure. The further first-principles calculations suggest that Li$_2$AuH$_6$ shows a high superconducting transition temperature ($T_c$) $\sim$ 140 K under ambient pressure. The H-1$s$ electrons strongly couple with phonon modes of vibrations of Au-H octahedrons as well as vibrations of Li atoms, where the latter is not taken seriously in other previously similar cases. Hence, different from previous claims of searching metallic covalent bonds to find high-$T_c$ superconductors, we emphasize here the importance of those phonon modes with strong electron-phonon coupling (EPC). And we suggest that one can intercalate atoms into binary or ternary hydrides to introduce more potential phonon modes with strong EPC, which is an effective approach to find high-$T_c$ superconductors within multicomponent compounds.
2501.12224
TokenVerse: Versatile Multi-concept Personalization in Token Modulation Space
cs.CV
We present TokenVerse -- a method for multi-concept personalization, leveraging a pre-trained text-to-image diffusion model. Our framework can disentangle complex visual elements and attributes from as little as a single image, while enabling seamless plug-and-play generation of combinations of concepts extracted from multiple images. As opposed to existing works, TokenVerse can handle multiple images with multiple concepts each, and supports a wide-range of concepts, including objects, accessories, materials, pose, and lighting. Our work exploits a DiT-based text-to-image model, in which the input text affects the generation through both attention and modulation (shift and scale). We observe that the modulation space is semantic and enables localized control over complex concepts. Building on this insight, we devise an optimization-based framework that takes as input an image and a text description, and finds for each word a distinct direction in the modulation space. These directions can then be used to generate new images that combine the learned concepts in a desired configuration. We demonstrate the effectiveness of TokenVerse in challenging personalization settings, and showcase its advantages over existing methods. project's webpage in https://token-verse.github.io/
2501.12226
CDW-CoT: Clustered Distance-Weighted Chain-of-Thoughts Reasoning
cs.LG
Large Language Models (LLMs) have recently achieved impressive results in complex reasoning tasks through Chain of Thought (CoT) prompting. However, most existing CoT methods rely on using the same prompts, whether manually designed or automatically generated, to handle the entire dataset. This one-size-fits-all approach may fail to meet the specific needs arising from the diversities within a single dataset. To solve this problem, we propose the Clustered Distance-Weighted Chain of Thought (CDW-CoT) method, which dynamically constructs prompts tailored to the characteristics of each data instance by integrating clustering and prompt optimization techniques. Our method employs clustering algorithms to categorize the dataset into distinct groups, from which a candidate pool of prompts is selected to reflect the inherent diversity within the dataset. For each cluster, CDW-CoT trains the optimal prompt probability distribution tailored to their specific characteristics. Finally, it dynamically constructs a unique prompt probability distribution for each test instance, based on its proximity to cluster centers, from which prompts are selected for reasoning. CDW-CoT consistently outperforms traditional CoT methods across six datasets, including commonsense, symbolic, and mathematical reasoning tasks. Specifically, when compared to manual CoT, CDW-CoT achieves an average accuracy improvement of 25.34% on LLaMA2 (13B) and 15.72% on LLaMA3 (8B).
2501.12227
Multi-terminal Strong Coordination over Noisy Channels with Encoder Cooperation
cs.IT math.IT
We investigate the problem of strong coordination over a multiple-access channel (MAC) with cribbing encoders. In this configuration, two encoders observe independent and identically distributed (i.i.d.) samples of a source random variable each and encode the inputs to the MAC. The decoder which observes the output of the MAC together with side-information, must generate approximately i.i.d. samples of another random variable which is jointly distributed with the two sources and the side information. We also allow for possible encoder cooperation, where one of the encoders can non-causally crib from the other encoders input. Independent pairwise shared randomness is assumed between each encoder and the decoder at limited rates. Firstly, in the presence of cribbing, we derive an achievable region based on joint source-channel coding. We also prove that in the absence of cribbing, our inner bound is tight for the special case when the MAC is composed of deterministic links, and the sources are conditionally independent given the side information. We then explicitly compute the regions for an example both with and without cribbing between the encoders, and demonstrate that cribbing strictly improves upon the achievable region.
2501.12231
InsTALL: Context-aware Instructional Task Assistance with Multi-modal Large Language Models
cs.CV cs.AI cs.CL
The improved competence of generative models can help building multi-modal virtual assistants that leverage modalities beyond language. By observing humans performing multi-step tasks, one can build assistants that have situational awareness of actions and tasks being performed, enabling them to cater assistance based on this understanding. In this paper, we develop a Context-aware Instructional Task Assistant with Multi-modal Large Language Models (InsTALL) that leverages an online visual stream (e.g. a user's screen share or video recording) and responds in real-time to user queries related to the task at hand. To enable useful assistance, InsTALL 1) trains a multi-modal model on task videos and paired textual data, and 2) automatically extracts task graph from video data and leverages it at training and inference time. We show InsTALL achieves state-of-the-art performance across proposed sub-tasks considered for multimodal activity understanding -- task recognition (TR), action recognition (AR), next action prediction (AP), and plan prediction (PP) -- and outperforms existing baselines on two novel sub-tasks related to automatic error identification.
2501.12234
Multi-Agent Feedback Motion Planning using Probably Approximately Correct Nonlinear Model Predictive Control
cs.RO
For many tasks, multi-robot teams often provide greater efficiency, robustness, and resiliency. However, multi-robot collaboration in real-world scenarios poses a number of major challenges, especially when dynamic robots must balance competing objectives like formation control and obstacle avoidance in the presence of stochastic dynamics and sensor uncertainty. In this paper, we propose a distributed, multi-agent receding-horizon feedback motion planning approach using Probably Approximately Correct Nonlinear Model Predictive Control (PAC-NMPC) that is able to reason about both model and measurement uncertainty to achieve robust multi-agent formation control while navigating cluttered obstacle fields and avoiding inter-robot collisions. Our approach relies not only on the underlying PAC-NMPC algorithm but also on a terminal cost-function derived from gyroscopic obstacle avoidance. Through numerical simulation, we show that our distributed approach performs on par with a centralized formulation, that it offers improved performance in the case of significant measurement noise, and that it can scale to more complex dynamical systems.
2501.12235
DLEN: Dual Branch of Transformer for Low-Light Image Enhancement in Dual Domains
cs.CV eess.IV
Low-light image enhancement (LLE) aims to improve the visual quality of images captured in poorly lit conditions, which often suffer from low brightness, low contrast, noise, and color distortions. These issues hinder the performance of computer vision tasks such as object detection, facial recognition, and autonomous driving.Traditional enhancement techniques, such as multi-scale fusion and histogram equalization, fail to preserve fine details and often struggle with maintaining the natural appearance of enhanced images under complex lighting conditions. Although the Retinex theory provides a foundation for image decomposition, it often amplifies noise, leading to suboptimal image quality. In this paper, we propose the Dual Light Enhance Network (DLEN), a novel architecture that incorporates two distinct attention mechanisms, considering both spatial and frequency domains. Our model introduces a learnable wavelet transform module in the illumination estimation phase, preserving high- and low-frequency components to enhance edge and texture details. Additionally, we design a dual-branch structure that leverages the power of the Transformer architecture to enhance both the illumination and structural components of the image.Through extensive experiments, our model outperforms state-of-the-art methods on standard benchmarks.Code is available here: https://github.com/LaLaLoXX/DLEN
2501.12236
Fast sparse optimization via adaptive shrinkage
math.OC cs.LG cs.SY eess.SY
The need for fast sparse optimization is emerging, e.g., to deal with large-dimensional data-driven problems and to track time-varying systems. In the framework of linear sparse optimization, the iterative shrinkage-thresholding algorithm is a valuable method to solve Lasso, which is particularly appreciated for its ease of implementation. Nevertheless, it converges slowly. In this paper, we develop a proximal method, based on logarithmic regularization, which turns out to be an iterative shrinkage-thresholding algorithm with adaptive shrinkage hyperparameter. This adaptivity substantially enhances the trajectory of the algorithm, in a way that yields faster convergence, while keeping the simplicity of the original method. Our contribution is twofold: on the one hand, we derive and analyze the proposed algorithm; on the other hand, we validate its fast convergence via numerical experiments and we discuss the performance with respect to state-of-the-art algorithms.
2501.12239
Investigating Market Strength Prediction with CNNs on Candlestick Chart Images
cs.CV
This paper investigates predicting market strength solely from candlestick chart images to assist investment decisions. The core research problem is developing an effective computer vision-based model using raw candlestick visuals without time-series data. We specifically analyze the impact of incorporating candlestick patterns that were detected by YOLOv8. The study implements two approaches: pure CNN on chart images and a Decomposer architecture detecting patterns. Experiments utilize diverse financial datasets spanning stocks, cryptocurrencies, and forex assets. Key findings demonstrate candlestick patterns do not improve model performance over only image data in our research. The significance is illuminating limitations in candlestick image signals. Performance peaked at approximately 0.7 accuracy, below more complex time-series models. Outcomes reveal challenges in distilling sufficient predictive power from visual shapes alone, motivating the incorporation of other data modalities. This research clarifies how purely image-based models can inform trading while confirming patterns add little value over raw charts. Our content is endeavored to be delineated into distinct sections, each autonomously furnishing a unique contribution while maintaining cohesive linkage. Note that, the examples discussed herein are not limited to the scope, applicability, or knowledge outlined in the paper.
2501.12243
FOCUS: First Order Concentrated Updating Scheme
cs.LG cs.CL math.OC
Large language models (LLMs) demonstrate remarkable performance, and improving their pre-training process appears to be key to enhancing their capabilities further. Based on the documented success of Adam, learning rate decay, and weight decay, we hypothesize that the pre-training loss landscape features a narrowing valley structure. Through experiments with synthetic loss functions, we discover that when gradient query noise is high relative to the valley's sharpness, Adam's performance falls behind that of Signum because Adam reduces the effective step size too drastically. This observation led us to develop FOCUS, an optimizer that enhances Signum by incorporating attraction toward moving averaged parameters, allowing it to handle noise better while maintaining larger step sizes. In training GPT-2, FOCUS proves to be more stable than Signum and faster than Adam. These results suggest that gradient noise may be an underappreciated limiting factor in LLM training, and FOCUS offers promising solutions.
2501.12244
Zero-shot Bias Correction: Efficient MR Image Inhomogeneity Reduction Without Any Data
eess.IV cs.CV
In recent years, deep neural networks for image inhomogeneity reduction have shown promising results. However, current methods with (un)supervised solutions require preparing a training dataset, which is expensive and laborious for data collection. In this work, we demonstrate a novel zero-shot deep neural networks, which requires no data for pre-training and dedicated assumption of the bias field. The designed light-weight CNN enables an efficient zero-shot adaptation for bias-corrupted image correction. Our method provides a novel solution to mitigate the biased corrupted image as iterative homogeneity refinement, which therefore ensures the considered issue can be solved easier with stable convergence of zero-shot optimization. Extensive comparison on different datasets show that the proposed method performs better than current data-free N4 methods in both efficiency and accuracy.
2501.12245
Quality Enhancement of Radiographic X-ray Images by Interpretable Mapping
eess.IV cs.CV
X-ray imaging is the most widely used medical imaging modality. However, in the common practice, inconsistency in the initial presentation of X-ray images is a common complaint by radiologists. Different patient positions, patient habitus and scanning protocols can lead to differences in image presentations, e.g., differences in brightness and contrast globally or regionally. To compensate for this, additional work will be executed by clinical experts to adjust the images to the desired presentation, which can be time-consuming. Existing deep-learning-based end-to-end solutions can automatically correct images with promising performances. Nevertheless, these methods are hard to be interpreted and difficult to be understood by clinical experts. In this manuscript, a novel interpretable mapping method by deep learning is proposed, which automatically enhances the image brightness and contrast globally and locally. Meanwhile, because the model is inspired by the workflow of the brightness and contrast manipulation, it can provide interpretable pixel maps for explaining the motivation of image enhancement. The experiment on the clinical datasets show the proposed method can provide consistent brightness and contrast correction on X-ray images with accuracy of 24.75 dB PSNR and 0.8431 SSIM.
2501.12246
Video Deblurring by Sharpness Prior Detection and Edge Information
cs.CV
Video deblurring is essential task for autonomous driving, facial recognition, and security surveillance. Traditional methods directly estimate motion blur kernels, often introducing artifacts and leading to poor results. Recent approaches utilize the detection of sharp frames within video sequences to enhance deblurring. However, existing datasets rely on fixed number of sharp frames, which may be too restrictive for some applications and may introduce a bias during model training. To address these limitations and enhance domain adaptability, this work first introduces GoPro Random Sharp (GoProRS), a new dataset where the the frequency of sharp frames within the sequence is customizable, allowing more diverse training and testing scenarios. Furthermore, it presents a novel video deblurring model, called SPEINet, that integrates sharp frame features into blurry frame reconstruction through an attention-based encoder-decoder architecture, a lightweight yet robust sharp frame detection and an edge extraction phase. Extensive experimental results demonstrate that SPEINet outperforms state-of-the-art methods across multiple datasets, achieving an average of +3.2% PSNR improvement over recent techniques. Given such promising results, we believe that both the proposed model and dataset pave the way for future advancements in video deblurring based on the detection of sharp frames.
2501.12251
Solar Panel Selection using Extended WASPAS with Disc Intuitionistic Fuzzy Choquet Integral Operators: CASPAS Methodology
cs.IT math.IT
Renewable energy is crucial for addressing the growing energy demands of modern society while mitigating the adverse effects of climate change. Unlike fossil fuels, renewable energy sources such as solar, wind, hydro, geothermal, and biomass are abundant, sustainable, and environmentally friendly. This study focuses on addressing a critical challenge in renewable energy decision-making by developing a novel framework for optimal solar panel selection, a key component of sustainable energy solutions. Solar panel selection involves evaluating multiple interdependent criteria, such as efficiency, cost, durability, and environmental impact. Traditional multi-criteria decision-making (MCDM) methods often fail to account for the interdependencies among these criteria, leading to suboptimal outcomes. To overcome this limitation, the study introduces the Choquet Aggregated Sum Product Assessment (CASPAS) method, a Choquet integral-based MCDM approach that incorporates fuzzy measures to model interactions among criteria. CASPAS generalizes the Weighted Aggregated Sum Product Assessment (WASPAS) method, thereby enhancing decision-making accuracy and reliability. This study also introduces the concept of disc intuitionistic fuzzy set (D-IFS), a generalization of the concept of circular intuitionistic fuzzy set, which employ a radius function capable of assigning varying values to individual elements instead of relying on a fixed radius. Recognizing that traditional weighted aggregation operators neglect the interaction among criteria, this study proposes disc intuitionistic fuzzy Choquet integral operators by incorporating the concept of fuzzy measures, which are effective in modeling such interactions. The proposed method is applied to a renewable energy problem on selecting optimal solar panels.
2501.12254
Memory Storyboard: Leveraging Temporal Segmentation for Streaming Self-Supervised Learning from Egocentric Videos
cs.CV cs.LG
Self-supervised learning holds the promise to learn good representations from real-world continuous uncurated data streams. However, most existing works in visual self-supervised learning focus on static images or artificial data streams. Towards exploring a more realistic learning substrate, we investigate streaming self-supervised learning from long-form real-world egocentric video streams. Inspired by the event segmentation mechanism in human perception and memory, we propose "Memory Storyboard" that groups recent past frames into temporal segments for more effective summarization of the past visual streams for memory replay. To accommodate efficient temporal segmentation, we propose a two-tier memory hierarchy: the recent past is stored in a short-term memory, and the storyboard temporal segments are then transferred to a long-term memory. Experiments on real-world egocentric video datasets including SAYCam and KrishnaCam show that contrastive learning objectives on top of storyboard frames result in semantically meaningful representations which outperform those produced by state-of-the-art unsupervised continual learning methods.
2501.12255
HAC++: Towards 100X Compression of 3D Gaussian Splatting
cs.CV
3D Gaussian Splatting (3DGS) has emerged as a promising framework for novel view synthesis, boasting rapid rendering speed with high fidelity. However, the substantial Gaussians and their associated attributes necessitate effective compression techniques. Nevertheless, the sparse and unorganized nature of the point cloud of Gaussians (or anchors in our paper) presents challenges for compression. To achieve a compact size, we propose HAC++, which leverages the relationships between unorganized anchors and a structured hash grid, utilizing their mutual information for context modeling. Additionally, HAC++ captures intra-anchor contextual relationships to further enhance compression performance. To facilitate entropy coding, we utilize Gaussian distributions to precisely estimate the probability of each quantized attribute, where an adaptive quantization module is proposed to enable high-precision quantization of these attributes for improved fidelity restoration. Moreover, we incorporate an adaptive masking strategy to eliminate invalid Gaussians and anchors. Overall, HAC++ achieves a remarkable size reduction of over 100X compared to vanilla 3DGS when averaged on all datasets, while simultaneously improving fidelity. It also delivers more than 20X size reduction compared to Scaffold-GS. Our code is available at https://github.com/YihangChen-ee/HAC-plus.
2501.12256
Lie-Bracket Nash Equilibrium Seeking with Bounded Update Rates for Noncooperative Games
math.OC cs.SY eess.SY
This paper proposes a novel approach for local convergence to Nash equilibrium in quadratic noncooperative games based on a distributed Lie-bracket extremum seeking control scheme. This is the first instance of noncooperative games being tackled in a model-free fashion integrated with the extremum seeking method of bounded update rates. In particular, the stability analysis is carried out using Lie-bracket approximation and Lyapunov's direct method. We quantify the size of the ultimate small residual sets around the Nash equilibrium and illustrate the theoretical results numerically on an example in an oligopoly setting.
2501.12263
mmCooper: A Multi-agent Multi-stage Communication-efficient and Collaboration-robust Cooperative Perception Framework
cs.CV
Collaborative perception significantly enhances individual vehicle perception performance through the exchange of sensory information among agents. However, real-world deployment faces challenges due to bandwidth constraints and inevitable calibration errors during information exchange. To address these issues, we propose mmCooper, a novel multi-agent, multi-stage, communication-efficient, and collaboration-robust cooperative perception framework. Our framework leverages a multi-stage collaboration strategy that dynamically and adaptively balances intermediate- and late-stage information to share among agents, enhancing perceptual performance while maintaining communication efficiency. To support robust collaboration despite potential misalignments and calibration errors, our framework captures multi-scale contextual information for robust fusion in the intermediate stage and calibrates the received detection results to improve accuracy in the late stage. We validate the effectiveness of mmCooper through extensive experiments on real-world and simulated datasets. The results demonstrate the superiority of our proposed framework and the effectiveness of each component.
2501.12266
CBVLM: Training-free Explainable Concept-based Large Vision Language Models for Medical Image Classification
cs.CV cs.AI cs.CL
The main challenges limiting the adoption of deep learning-based solutions in medical workflows are the availability of annotated data and the lack of interpretability of such systems. Concept Bottleneck Models (CBMs) tackle the latter by constraining the final disease prediction on a set of predefined and human-interpretable concepts. However, the increased interpretability achieved through these concept-based explanations implies a higher annotation burden. Moreover, if a new concept needs to be added, the whole system needs to be retrained. Inspired by the remarkable performance shown by Large Vision-Language Models (LVLMs) in few-shot settings, we propose a simple, yet effective, methodology, CBVLM, which tackles both of the aforementioned challenges. First, for each concept, we prompt the LVLM to answer if the concept is present in the input image. Then, we ask the LVLM to classify the image based on the previous concept predictions. Moreover, in both stages, we incorporate a retrieval module responsible for selecting the best examples for in-context learning. By grounding the final diagnosis on the predicted concepts, we ensure explainability, and by leveraging the few-shot capabilities of LVLMs, we drastically lower the annotation cost. We validate our approach with extensive experiments across four medical datasets and twelve LVLMs (both generic and medical) and show that CBVLM consistently outperforms CBMs and task-specific supervised methods without requiring any training and using just a few annotated examples. More information on our project page: https://cristianopatricio.github.io/CBVLM/.
2501.12267
VipDiff: Towards Coherent and Diverse Video Inpainting via Training-free Denoising Diffusion Models
cs.CV
Recent video inpainting methods have achieved encouraging improvements by leveraging optical flow to guide pixel propagation from reference frames either in the image space or feature space. However, they would produce severe artifacts in the mask center when the masked area is too large and no pixel correspondences can be found for the center. Recently, diffusion models have demonstrated impressive performance in generating diverse and high-quality images, and have been exploited in a number of works for image inpainting. These methods, however, cannot be applied directly to videos to produce temporal-coherent inpainting results. In this paper, we propose a training-free framework, named VipDiff, for conditioning diffusion model on the reverse diffusion process to produce temporal-coherent inpainting results without requiring any training data or fine-tuning the pre-trained diffusion models. VipDiff takes optical flow as guidance to extract valid pixels from reference frames to serve as constraints in optimizing the randomly sampled Gaussian noise, and uses the generated results for further pixel propagation and conditional generation. VipDiff also allows for generating diverse video inpainting results over different sampled noise. Experiments demonstrate that VipDiff can largely outperform state-of-the-art video inpainting methods in terms of both spatial-temporal coherence and fidelity.
2501.12269
Benchmarking Image Perturbations for Testing Automated Driving Assistance Systems
cs.SE cs.CV
Advanced Driver Assistance Systems (ADAS) based on deep neural networks (DNNs) are widely used in autonomous vehicles for critical perception tasks such as object detection, semantic segmentation, and lane recognition. However, these systems are highly sensitive to input variations, such as noise and changes in lighting, which can compromise their effectiveness and potentially lead to safety-critical failures. This study offers a comprehensive empirical evaluation of image perturbations, techniques commonly used to assess the robustness of DNNs, to validate and improve the robustness and generalization of ADAS perception systems. We first conducted a systematic review of the literature, identifying 38 categories of perturbations. Next, we evaluated their effectiveness in revealing failures in two different ADAS, both at the component and at the system level. Finally, we explored the use of perturbation-based data augmentation and continuous learning strategies to improve ADAS adaptation to new operational design domains. Our results demonstrate that all categories of image perturbations successfully expose robustness issues in ADAS and that the use of dataset augmentation and continuous learning significantly improves ADAS performance in novel, unseen environments.
2501.12271
Faithful Simulation of Distributed Quantum Measurement with Coding for Computing
cs.IT math.IT
This papers consider a two terminal problem, where Alice and Bob jointly want to perform a measurement on a bipartite quantum system $\rho^{AB}$. Alice can transmit the results of her measurements to Bob on a classical channel, and Alice and Bob have common randomness. The question is what is the minimum amount of communications and common randomness needed for faithful simulation. The paper derives an achievable rate region.
2501.12272
A Lightweight Approach for User and Keyword Classification in Controversial Topics
cs.SI
Classifying the stance of individuals on controversial topics and uncovering their concerns is crucial for social scientists and policymakers. Data from Online Social Networks (OSNs), which serve as a proxy to a representative sample of society, offers an opportunity to classify these stances, discover society's concerns regarding controversial topics, and track the evolution of these concerns over time. Consequently, stance classification in OSNs has garnered significant attention from researchers. However, most existing methods for this task often rely on labelled data and utilise the text of users' posts or the interactions between users, necessitating large volumes of data, considerable processing time, and access to information that is not readily available (e.g. users' followers/followees). This paper proposes a lightweight approach for the stance classification of users and keywords in OSNs, aiming at understanding the collective opinion of individuals and their concerns. Our approach employs a tailored random walk model, requiring just one keyword representing each stance, using solely the keywords in social media posts. Experimental results demonstrate the superior performance of our method compared to the baselines, excelling in stance classification of users and keywords, with a running time that, while not the fastest, remains competitive.
2501.12273
Condor: Enhance LLM Alignment with Knowledge-Driven Data Synthesis and Refinement
cs.CL cs.AI
The quality of Supervised Fine-Tuning (SFT) data plays a critical role in enhancing the conversational capabilities of Large Language Models (LLMs). However, as LLMs become more advanced, the availability of high-quality human-annotated SFT data has become a significant bottleneck, necessitating a greater reliance on synthetic training data. In this work, we introduce Condor, a novel two-stage synthetic data generation framework that incorporates World Knowledge Tree and Self-Reflection Refinement to produce high-quality SFT data at scale. Our experimental results demonstrate that a base model fine-tuned on only 20K Condor-generated samples achieves superior performance compared to counterparts. The additional refinement stage in Condor further enables iterative self-improvement for LLMs at various scales (up to 72B), validating the effectiveness of our approach. Furthermore, our investigation into the scaling for synthetic data in post-training reveals substantial unexplored potential for performance improvements, opening promising avenues for future research.
2501.12274
Making it to First: The Random Access Problem in DNA Storage
cs.IT math.IT
We study the Random Access Problem in DNA storage, which addresses the challenge of retrieving a specific information strand from a DNA-based storage system. Given that $k$ information strands, representing the data, are encoded into $n$ strands using a code. The goal under this paradigm is to identify and analyze codes that minimize the expected number of reads required to retrieve any of the $k$ information strand, while in each read one of the $n$ encoded strands is read uniformly at random. We fully solve the case when $k=2$, showing that the best possible code attains a random access expectation of $0.914 \cdot 2$. Moreover, we generalize a construction from \cite{GMZ24}, specific to $k=3$, for any value of $k$. Our construction uses $B_{k-1}$ sequences over $\mathbb{Z}_{q-1}$, that always exist over large finite fields. For $k=4$, we show that this generalized construction outperforms all previous constructions in terms of reducing the random access expectation .
2501.12275
With Great Backbones Comes Great Adversarial Transferability
cs.CV cs.AI cs.CR cs.LG cs.MA
Advances in self-supervised learning (SSL) for machine vision have improved representation robustness and model performance, giving rise to pre-trained backbones like \emph{ResNet} and \emph{ViT} models tuned with SSL methods such as \emph{SimCLR}. Due to the computational and data demands of pre-training, the utilization of such backbones becomes a strenuous necessity. However, employing these backbones may inherit vulnerabilities to adversarial attacks. While adversarial robustness has been studied under \emph{white-box} and \emph{black-box} settings, the robustness of models tuned on pre-trained backbones remains largely unexplored. Additionally, the role of tuning meta-information in mitigating exploitation risks is unclear. This work systematically evaluates the adversarial robustness of such models across $20,000$ combinations of tuning meta-information, including fine-tuning techniques, backbone families, datasets, and attack types. We propose using proxy models to transfer attacks, simulating varying levels of target knowledge by fine-tuning these proxies with diverse configurations. Our findings reveal that proxy-based attacks approach the effectiveness of \emph{white-box} methods, even with minimal tuning knowledge. We also introduce a naive "backbone attack," leveraging only the backbone to generate adversarial samples, which outperforms \emph{black-box} attacks and rivals \emph{white-box} methods, highlighting critical risks in model-sharing practices. Finally, our ablations reveal how increasing tuning meta-information impacts attack transferability, measuring each meta-information combination.
2501.12279
Spatial exponential decay of perturbations in optimal control of general evolution equations
math.OC cs.SY eess.SY math.AP
We analyze the robustness of optimally controlled evolution equations with respect to spatially localized perturbations. We prove that if the involved operators are domain-uniformly stabilizable and detectable, then these localized perturbations only have a local effect on the optimal solution. We characterize this domain-uniform stabilizability and detectability for the transport equation with constant transport velocity, showing that even for unitary semigroups, optimality implies exponential damping. Finally, we extend our result to the case of a space-dependent transport velocity. Numerical examples in one space dimension complement the theoretical results.
2501.12280
Bounds and Codes for General Phased Burst Errors
cs.IT math.IT
Phased burst errors (PBEs) are bursts of errors occurring at one or more known locations. The correction of PBEs is a classical topic in coding theory, with prominent applications such as the design of array codes for memory systems or distributed storage. We propose a general yet fine-grained approach to this problem, accounting not only for the number of bursts but also the error structure in each burst. By modeling PBEs as an error set in an adversarial channel, we investigate bounds on the maximal size of codes that can correct them. The PBE-correction capability of generalized concatenated codes is analyzed, and asymptotically good PBE-correcting codes are constructed, recovering a classical construction in a specific problem instance.
2501.12281
MoGERNN: An Inductive Traffic Predictor for Unobserved Locations in Dynamic Sensing Networks
cs.LG
Given a partially observed road network, how can we predict the traffic state of unobserved locations? While deep learning approaches show exceptional performance in traffic prediction, most assume sensors at all locations of interest, which is impractical due to financial constraints. Furthermore, these methods typically require costly retraining when sensor configurations change. We propose MoGERNN, an inductive spatio-temporal graph representation model, to address these challenges. Inspired by the Mixture of Experts approach in Large Language Models, we introduce a Mixture of Graph Expert (MoGE) block to model complex spatial dependencies through multiple graph message aggregators and a sparse gating network. This block estimates initial states for unobserved locations, which are then processed by a GRU-based Encoder-Decoder that integrates a graph message aggregator to capture spatio-temporal dependencies and predict future states. Experiments on two real-world datasets show MoGERNN consistently outperforms baseline methods for both observed and unobserved locations. MoGERNN can accurately predict congestion evolution even in areas without sensors, offering valuable information for traffic management. Moreover, MoGERNN is adaptable to dynamic sensing networks, maintaining competitive performance even compared to its retrained counterpart. Tests with different numbers of available sensors confirm its consistent superiority, and ablation studies validate the effectiveness of its key modules.
2501.12285
Implementation of an Asymmetric Adjusted Activation Function for Class Imbalance Credit Scoring
cs.LG cs.AI q-fin.RM
Credit scoring is a systematic approach to evaluate a borrower's probability of default (PD) on a bank loan. The data associated with such scenarios are characteristically imbalanced, complicating binary classification owing to the often-underestimated cost of misclassification during the classifier's learning process. Considering the high imbalance ratio (IR) of these datasets, we introduce an innovative yet straightforward optimized activation function by incorporating an IR-dependent asymmetric adjusted factor embedded Sigmoid activation function (ASIG). The embedding of ASIG makes the sensitive margin of the Sigmoid function auto-adjustable, depending on the imbalance nature of the datasets distributed, thereby giving the activation function an asymmetric characteristic that prevents the underrepresentation of the minority class (positive samples) during the classifier's learning process. The experimental results show that the ASIG-embedded-classifier outperforms traditional classifiers on datasets across wide-ranging IRs in the downstream credit-scoring task. The algorithm also shows robustness and stability, even when the IR is ultra-high. Therefore, the algorithm provides a competitive alternative in the financial industry, especially in credit scoring, possessing the ability to effectively process highly imbalanced distribution data.
2501.12286
A Linear Programming Approach to Private Information Retrieval
cs.IT math.IT
This work presents an algorithmic framework that uses linear programming to construct \emph{addition-based Private Information Retrieval (AB-PIR)} schemes, where retrieval is performed by downloading only linear combinations of message symbols with coefficients set to 0 or 1. The AB-PIR schemes generalize several existing capacity-achieving PIR schemes and are of practical interest because they use only addition operations -- avoiding multiplication and other complex operations -- and are compatible with any finite field, including binary. Our framework broadens the search space to include all feasible solutions and can be used to construct optimal AB-PIR schemes for the entire range of problem parameters, including the number of servers, the total number of messages, and the number of messages that need to be retrieved. The framework enables us to identify schemes that outperform the previously proposed PIR schemes in certain cases and, in other cases, achieve performance on par with the best-known AB-PIR solutions. Additionally, the schemes generated by our framework can be integrated into existing solutions for several related PIR scenarios, improving their overall performance.
2501.12288
Microgrid Operation Control with State-of-Charge- Dependent Storage Power Constraints
eess.SY cs.SY
The microgrid concept offers high flexibility and resilience due to the possibility of switching between grid-connected and stand-alone operation. This renders microgrids an auspicious solution for rural areas and critical infrastructure. In standalone or islanded mode, the main objective is cost minimization while ensuring a safe and reliable operation. Optimal operation schemes for microgrids usually assume fixed power limits for energy storage units. This, however, is not sufficient for lithiumion energy storage systems, which often come with dynamic power limits that depend on the state of charge. These limits are especially prominent when the state of charge is close to its boundaries. In this paper, dynamic constraints for energy storages are modelled using convex polytopes and fitted to experimental data acquired from an 11.6 kWh lithium-ion energy storage system. The polytopic constraints are integrated in a model predictive control scheme that was designed for a standalone microgrid composed of a fuel cell, a photovoltaic generator and a lithium-ion energy storage system. To evaluate the advantages, a case study with two configurations is performed. The model predictive controller without polytopic constraints led to constraint violations in 11.77 % of the simulation time steps with a maximum deviation of 118 % above the power limits. The configuration with polytopic constraints in contrary led to no violations over the entire simulation horizon.
2501.12289
Regressor-Guided Image Editing Regulates Emotional Response to Reduce Online Engagement
cs.CV cs.AI cs.HC
Emotions are known to mediate the relationship between users' content consumption and their online engagement, with heightened emotional intensity leading to increased engagement. Building on this insight, we propose three regressor-guided image editing approaches aimed at diminishing the emotional impact of images. These include (i) a parameter optimization approach based on global image transformations known to influence emotions, (ii) an optimization approach targeting the style latent space of a generative adversarial network, and (iii) a diffusion-based approach employing classifier guidance and classifier-free guidance. Our findings demonstrate that approaches can effectively alter the emotional properties of images while maintaining high visual quality. Optimization-based methods primarily adjust low-level properties like color hues and brightness, whereas the diffusion-based approach introduces semantic changes, such as altering appearance or facial expressions. Notably, results from a behavioral study reveal that only the diffusion-based approach successfully elicits changes in viewers' emotional responses while preserving high perceived image quality. In future work, we will investigate the impact of these image adaptations on internet user behavior.
2501.12293
Improved Decoding of Tanner Codes
cs.IT cs.CC cs.DS math.IT
In this paper, we present improved decoding algorithms for expander-based Tanner codes. We begin by developing a randomized linear-time decoding algorithm that, under the condition that $ \delta d_0 > 2 $, corrects up to $ \alpha n $ errors for a Tanner code $ T(G, C_0) $, where $ G $ is a $ (c, d, \alpha, \delta) $-bipartite expander with $n$ left vertices, and $ C_0 \subseteq \mathbb{F}_2^d $ is a linear inner code with minimum distance $ d_0 $. This result improves upon the previous work of Cheng, Ouyang, Shangguan, and Shen (RANDOM 2024), which required $ \delta d_0 > 3 $. We further derandomize the algorithm to obtain a deterministic linear-time decoding algorithm with the same decoding radius. Our algorithm improves upon the previous deterministic algorithm of Cheng et al.\ by achieving a decoding radius of $ \alpha n $, compared with the previous radius of $ \frac{2\alpha}{d_0(1 + 0.5c\delta) }n$. Additionally, we investigate the size-expansion trade-off introduced by the recent work of Chen, Cheng, Li, and Ouyang (IEEE TIT 2023), and use it to provide new bounds on the minimum distance of Tanner codes. Specifically, we prove that the minimum distance of a Tanner code $T(G,C_0)$ is approximately $f_\delta^{-1} \left( \frac{1}{d_0} \right) \alpha n $, where $ f_\delta(\cdot) $ is the Size-Expansion Function. As another application, we improve the decoding radius of our decoding algorithms from $\alpha n$ to approximately $f_\delta^{-1}(\frac{2}{d_0})\alpha n$.
2501.12294
Wrap-Decoding in Asynchronous Unsourced Multiple Access With and Without Delay Information
cs.IT math.IT
An asynchronous $\ka$-active-user unsourced multiple access channel (AUMAC) is a key model for uncoordinated massive access in future networks. We focus on a scenario where each transmission is subject to the maximal delay constraint ($\dm$), and the precise delay of each user is unknown at the receiver. The combined effects of asynchronicity and uncertain delays require analysis over all possible delay-codeword combinations, making the complexity of the analysis grow with $\dm$ and $\ka$ exponentially. To overcome the complexity, we employ a wrap-decoder for the AUMAC and derive a uniform upper bound on the per-user probability of error (PUPE). The numerical result shows the trade-off between energy per bit and the number of active users under various delay constraints. Furthermore, in our considered AUMAC, decoding without explicit delay information is shown to achieve nearly the same energy efficiency as decoding with perfect delay knowledge.
2501.12295
Towards Accurate Unified Anomaly Segmentation
cs.CV
Unsupervised anomaly detection (UAD) from images strives to model normal data distributions, creating discriminative representations to distinguish and precisely localize anomalies. Despite recent advancements in the efficient and unified one-for-all scheme, challenges persist in accurately segmenting anomalies for further monitoring. Moreover, this problem is obscured by the widely-used AUROC metric under imbalanced UAD settings. This motivates us to emphasize the significance of precise segmentation of anomaly pixels using pAP and DSC as metrics. To address the unsolved segmentation task, we introduce the Unified Anomaly Segmentation (UniAS). UniAS presents a multi-level hybrid pipeline that progressively enhances normal information from coarse to fine, incorporating a novel multi-granularity gated CNN (MGG-CNN) into Transformer layers to explicitly aggregate local details from different granularities. UniAS achieves state-of-the-art anomaly segmentation performance, attaining 65.12/59.33 and 40.06/32.50 in pAP/DSC on the MVTec-AD and VisA datasets, respectively, surpassing previous methods significantly. The codes are shared at https://github.com/Mwxinnn/UniAS.
2501.12296
RALAD: Bridging the Real-to-Sim Domain Gap in Autonomous Driving with Retrieval-Augmented Learning
cs.CV cs.AI
In the pursuit of robust autonomous driving systems, models trained on real-world datasets often struggle to adapt to new environments, particularly when confronted with corner cases such as extreme weather conditions. Collecting these corner cases in the real world is non-trivial, which necessitates the use of simulators for validation. However,the high computational cost and the domain gap in data distribution have hindered the seamless transition between real and simulated driving scenarios. To tackle this challenge, we propose Retrieval-Augmented Learning for Autonomous Driving (RALAD), a novel framework designed to bridge the real-to-sim gap at a low cost. RALAD features three primary designs, including (1) domain adaptation via an enhanced Optimal Transport (OT) method that accounts for both individual and grouped image distances, (2) a simple and unified framework that can be applied to various models, and (3) efficient fine-tuning techniques that freeze the computationally expensive layers while maintaining robustness. Experimental results demonstrate that RALAD compensates for the performance degradation in simulated environments while maintaining accuracy in real-world scenarios across three different models. Taking Cross View as an example, the mIOU and mAP metrics in real-world scenarios remain stable before and after RALAD fine-tuning, while in simulated environments,the mIOU and mAP metrics are improved by 10.30% and 12.29%, respectively. Moreover, the re-training cost of our approach is reduced by approximately 88.1%. Our code is available at https://github.com/JiachengZuo/RALAD.git.
2501.12299
Sublinear Variational Optimization of Gaussian Mixture Models with Millions to Billions of Parameters
stat.ML cs.CV cs.LG
Gaussian Mixture Models (GMMs) range among the most frequently used machine learning models. However, training large, general GMMs becomes computationally prohibitive for datasets with many data points $N$ of high-dimensionality $D$. For GMMs with arbitrary covariances, we here derive a highly efficient variational approximation, which is integrated with mixtures of factor analyzers (MFAs). For GMMs with $C$ components, our proposed algorithm significantly reduces runtime complexity per iteration from $\mathcal{O}(NCD^2)$ to a complexity scaling linearly with $D$ and remaining constant w.r.t. $C$. Numerical validation of this theoretical complexity reduction then shows the following: the distance evaluations required for the entire GMM optimization process scale sublinearly with $NC$. On large-scale benchmarks, this sublinearity results in speed-ups of an order-of-magnitude compared to the state-of-the-art. As a proof of concept, we train GMMs with over 10 billion parameters on about 100 million images, and observe training times of approximately nine hours on a single state-of-the-art CPU.
2501.12300
LLM-Assisted Knowledge Graph Completion for Curriculum and Domain Modelling in Personalized Higher Education Recommendations
cs.HC cs.AI
While learning personalization offers great potential for learners, modern practices in higher education require a deeper consideration of domain models and learning contexts, to develop effective personalization algorithms. This paper introduces an innovative approach to higher education curriculum modelling that utilizes large language models (LLMs) for knowledge graph (KG) completion, with the goal of creating personalized learning-path recommendations. Our research focuses on modelling university subjects and linking their topics to corresponding domain models, enabling the integration of learning modules from different faculties and institutions in the student's learning path. Central to our approach is a collaborative process, where LLMs assist human experts in extracting high-quality, fine-grained topics from lecture materials. We develop a domain, curriculum, and user models for university modules and stakeholders. We implement this model to create the KG from two study modules: Embedded Systems and Development of Embedded Systems Using FPGA. The resulting KG structures the curriculum and links it to the domain models. We evaluate our approach through qualitative expert feedback and quantitative graph quality metrics. Domain experts validated the relevance and accuracy of the model, while the graph quality metrics measured the structural properties of our KG. Our results show that the LLM-assisted graph completion approach enhances the ability to connect related courses across disciplines to personalize the learning experience. Expert feedback also showed high acceptance of the proposed collaborative approach for concept extraction and classification.
2501.12309
A Hybrid Supervised and Self-Supervised Graph Neural Network for Edge-Centric Applications
cs.LG q-bio.MN
This paper presents a novel graph-based deep learning model for tasks involving relations between two nodes (edge-centric tasks), where the focus lies on predicting relationships and interactions between pairs of nodes rather than node properties themselves. This model combines supervised and self-supervised learning, taking into account for the loss function the embeddings learned and patterns with and without ground truth. Additionally it incorporates an attention mechanism that leverages both node and edge features. The architecture, trained end-to-end, comprises two primary components: embedding generation and prediction. First, a graph neural network (GNN) transform raw node features into dense, low-dimensional embeddings, incorporating edge attributes. Then, a feedforward neural model processes the node embeddings to produce the final output. Experiments demonstrate that our model matches or exceeds existing methods for protein-protein interactions prediction and Gene Ontology (GO) terms prediction. The model also performs effectively with one-hot encoding for node features, providing a solution for the previously unsolved problem of predicting similarity between compounds with unknown structures.
2501.12310
Optimizing Leaky Private Information Retrieval Codes to Achieve ${O}(\log K)$ Leakage Ratio Exponent
cs.IR cs.IT math.IT
We study the problem of leaky private information retrieval (L-PIR), where the amount of privacy leakage is measured by the pure differential privacy parameter, referred to as the leakage ratio exponent. Unlike the previous L-PIR scheme proposed by Samy et al., which only adjusted the probability allocation to the clean (low-cost) retrieval pattern, we optimize the probabilities assigned to all the retrieval patterns jointly. It is demonstrated that the optimal retrieval pattern probability distribution is quite sophisticated and has a layered structure: the retrieval patterns associated with the random key values of lower Hamming weights should be assigned higher probabilities. This new scheme provides a significant improvement, leading to an ${O}(\log K)$ leakage ratio exponent with fixed download cost $D$ and number of servers $N$, in contrast to the previous art that only achieves a $\Theta(K)$ exponent, where $K$ is the number of messages.
2501.12314
Uncertainty Quantification With Noise Injection in Neural Networks: A Bayesian Perspective
stat.ML cs.LG
Model uncertainty quantification involves measuring and evaluating the uncertainty linked to a model's predictions, helping assess their reliability and confidence. Noise injection is a technique used to enhance the robustness of neural networks by introducing randomness. In this paper, we establish a connection between noise injection and uncertainty quantification from a Bayesian standpoint. We theoretically demonstrate that injecting noise into the weights of a neural network is equivalent to Bayesian inference on a deep Gaussian process. Consequently, we introduce a Monte Carlo Noise Injection (MCNI) method, which involves injecting noise into the parameters during training and performing multiple forward propagations during inference to estimate the uncertainty of the prediction. Through simulation and experiments on regression and classification tasks, our method demonstrates superior performance compared to the baseline model.
2501.12318
BlanketGen2-Fit3D: Synthetic Blanket Augmentation Towards Improving Real-World In-Bed Blanket Occluded Human Pose Estimation
cs.CV
Human Pose Estimation (HPE) from monocular RGB images is crucial for clinical in-bed skeleton-based action recognition, however, it poses unique challenges for HPE models due to the frequent presence of blankets occluding the person, while labeled HPE data in this scenario is scarce. To address this we introduce BlanketGen2-Fit3D (BG2-Fit3D), an augmentation of Fit3D dataset that contains 1,217,312 frames with synthetic photo-realistic blankets. To generate it we used BlanketGen2, our new and improved version of our BlanketGen pipeline that simulates synthetic blankets using ground-truth Skinned Multi-Person Linear model (SMPL) meshes and then renders them as transparent images that can be layered on top of the original frames. This dataset was used in combination with the original Fit3D to finetune the ViTPose-B HPE model, to evaluate synthetic blanket augmentation effectiveness. The trained models were further evaluated on a real-world blanket occluded in-bed HPE dataset (SLP dataset). Comparing architectures trained on only Fit3D with the ones trained with our synthetic blanket augmentation the later improved pose estimation performance on BG2-Fit3D, the synthetic blanket occluded dataset significantly to (0.977 Percentage of Correct Keypoints (PCK), 0.149 Normalized Mean Error (NME)) with an absolute 4.4% PCK increase. Furthermore, the test results on SLP demonstrated the utility of synthetic data augmentation by improving performance by an absolute 2.3% PCK, on real-world images with the poses occluded by real blankets. These results show synthetic blanket augmentation has the potential to improve in-bed blanket occluded HPE from RGB images. The dataset as well as the code will be made available to the public.
2501.12319
Metric for Evaluating Performance of Reference-Free Demorphing Methods
cs.CV
A facial morph is an image created by combining two (or more) face images pertaining to two (or more) distinct identities. Reference-free face demorphing inverts the process and tries to recover the face images constituting a facial morph without using any other information. However, there is no consensus on the evaluation metrics to be used to evaluate and compare such demorphing techniques. In this paper, we first analyze the shortcomings of the demorphing metrics currently used in the literature. We then propose a new metric called biometrically cross-weighted IQA that overcomes these issues and extensively benchmark current methods on the proposed metric to show its efficacy. Experiments on three existing demorphing methods and six datasets on two commonly used face matchers validate the efficacy of our proposed metric.
2501.12322
A General Achievable Scheme for Linear Computation Broadcast Channel
cs.IT math.IT
This paper presents a new achievable scheme for the Linear Computation Broadcast Channel (LCBC), which is based on a generalized subspace decomposition derived from representable polymatroid space. This decomposition enables the server to serve user demands with an approach of effective multicast and interference elimination. We extend existing results by introducing a linear programming framework to optimize multicast opportunities across an arbitrary number of users.
2501.12323
Deep Learning Based Segmentation of Blood Vessels from H&E Stained Oesophageal Adenocarcinoma Whole-Slide Images
eess.IV cs.CV
Blood vessels (BVs) play a critical role in the Tumor Micro-Environment (TME), potentially influencing cancer progression and treatment response. However, manually quantifying BVs in Hematoxylin and Eosin (H&E) stained images is challenging and labor-intensive due to their heterogeneous appearances. We propose a novel approach of constructing guiding maps to improve the performance of state-of-the-art segmentation models for BV segmentation, the guiding maps encourage the models to learn representative features of BVs. This is particularly beneficial for computational pathology, where labeled training data is often limited and large models are prone to overfitting. We have quantitative and qualitative results to demonstrate the efficacy of our approach in improving segmentation accuracy. In future, we plan to validate this method to segment BVs across various tissue types and investigate the role of cellular structures in relation to BVs in the TME.
2501.12326
UI-TARS: Pioneering Automated GUI Interaction with Native Agents
cs.AI cs.CL cs.CV cs.HC
This paper introduces UI-TARS, a native GUI agent model that solely perceives the screenshots as input and performs human-like interactions (e.g., keyboard and mouse operations). Unlike prevailing agent frameworks that depend on heavily wrapped commercial models (e.g., GPT-4o) with expert-crafted prompts and workflows, UI-TARS is an end-to-end model that outperforms these sophisticated frameworks. Experiments demonstrate its superior performance: UI-TARS achieves SOTA performance in 10+ GUI agent benchmarks evaluating perception, grounding, and GUI task execution. Notably, in the OSWorld benchmark, UI-TARS achieves scores of 24.6 with 50 steps and 22.7 with 15 steps, outperforming Claude (22.0 and 14.9 respectively). In AndroidWorld, UI-TARS achieves 46.6, surpassing GPT-4o (34.5). UI-TARS incorporates several key innovations: (1) Enhanced Perception: leveraging a large-scale dataset of GUI screenshots for context-aware understanding of UI elements and precise captioning; (2) Unified Action Modeling, which standardizes actions into a unified space across platforms and achieves precise grounding and interaction through large-scale action traces; (3) System-2 Reasoning, which incorporates deliberate reasoning into multi-step decision making, involving multiple reasoning patterns such as task decomposition, reflection thinking, milestone recognition, etc. (4) Iterative Training with Reflective Online Traces, which addresses the data bottleneck by automatically collecting, filtering, and reflectively refining new interaction traces on hundreds of virtual machines. Through iterative training and reflection tuning, UI-TARS continuously learns from its mistakes and adapts to unforeseen situations with minimal human intervention. We also analyze the evolution path of GUI agents to guide the further development of this domain.
2501.12327
VARGPT: Unified Understanding and Generation in a Visual Autoregressive Multimodal Large Language Model
cs.CV
We present VARGPT, a novel multimodal large language model (MLLM) that unifies visual understanding and generation within a single autoregressive framework. VARGPT employs a next-token prediction paradigm for visual understanding and a next-scale prediction paradigm for visual autoregressive generation. VARGPT innovatively extends the LLaVA architecture, achieving efficient scale-wise autoregressive visual generation within MLLMs while seamlessly accommodating mixed-modal input and output within a single model framework. Our VARGPT undergoes a three-stage unified training process on specially curated datasets, comprising a pre-training phase and two mixed visual instruction-tuning phases. The unified training strategy are designed to achieve alignment between visual and textual features, enhance instruction following for both understanding and generation, and improve visual generation quality, respectively. Despite its LLAVA-based architecture for multimodel understanding, VARGPT significantly outperforms LLaVA-1.5 across various vision-centric benchmarks, such as visual question-answering and reasoning tasks. Notably, VARGPT naturally supports capabilities in autoregressive visual generation and instruction-to-image synthesis, showcasing its versatility in both visual understanding and generation tasks. Project page is at: \url{https://vargpt-1.github.io/}
2501.12330
The Gap Between Principle and Practice of Lossy Image Coding
cs.IT cs.LG math.IT
Lossy image coding is the art of computing that is principally bounded by the image's rate-distortion function. This bound, though never accurately characterized, has been approached practically via deep learning technologies in recent years. Indeed, learned image coding schemes allow direct optimization of the joint rate-distortion cost, thereby outperforming the handcrafted image coding schemes by a large margin. Still, it is observed that there is room for further improvement in the rate-distortion performance of learned image coding. In this article, we identify the gap between the ideal rate-distortion function forecasted by Shannon's information theory and the empirical rate-distortion function achieved by the state-of-the-art learned image coding schemes, revealing that the gap is incurred by five different effects: modeling effect, approximation effect, amortization effect, digitization effect, and asymptotic effect. We design simulations and experiments to quantitively evaluate the last three effects, which demonstrates the high potential of future lossy image coding technologies.
2501.12331
Cinepro: Robust Training of Foundation Models for Cancer Detection in Prostate Ultrasound Cineloops
eess.IV cs.CV cs.LG q-bio.TO
Prostate cancer (PCa) detection using deep learning (DL) models has shown potential for enhancing real-time guidance during biopsies. However, prostate ultrasound images lack pixel-level cancer annotations, introducing label noise. Current approaches often focus on limited regions of interest (ROIs), disregarding anatomical context necessary for accurate diagnosis. Foundation models can overcome this limitation by analyzing entire images to capture global spatial relationships; however, they still encounter challenges stemming from the weak labels associated with coarse pathology annotations in ultrasound data. We introduce Cinepro, a novel framework that strengthens foundation models' ability to localize PCa in ultrasound cineloops. Cinepro adapts robust training by integrating the proportion of cancer tissue reported by pathology in a biopsy core into its loss function to address label noise, providing a more nuanced supervision. Additionally, it leverages temporal data across multiple frames to apply robust augmentations, enhancing the model's ability to learn stable cancer-related features. Cinepro demonstrates superior performance on a multi-center prostate ultrasound dataset, achieving an AUROC of 77.1% and a balanced accuracy of 83.8%, surpassing current benchmarks. These findings underscore Cinepro's promise in advancing foundation models for weakly labeled ultrasound data.
2501.12332
Automatic Labelling with Open-source LLMs using Dynamic Label Schema Integration
cs.CL cs.AI cs.LG
Acquiring labelled training data remains a costly task in real world machine learning projects to meet quantity and quality requirements. Recently Large Language Models (LLMs), notably GPT-4, have shown great promises in labelling data with high accuracy. However, privacy and cost concerns prevent the ubiquitous use of GPT-4. In this work, we explore effectively leveraging open-source models for automatic labelling. We identify integrating label schema as a promising technology but found that naively using the label description for classification leads to poor performance on high cardinality tasks. To address this, we propose Retrieval Augmented Classification (RAC) for which LLM performs inferences for one label at a time using corresponding label schema; we start with the most related label and iterates until a label is chosen by the LLM. We show that our method, which dynamically integrates label description, leads to performance improvements in labelling tasks. We further show that by focusing only on the most promising labels, RAC can trade off between label quality and coverage - a property we leverage to automatically label our internal datasets.
2501.12336
FuocChuVIP123 at CoMeDi Shared Task: Disagreement Ranking with XLM-Roberta Sentence Embeddings and Deep Neural Regression
cs.CL cs.AI
This paper presents results of our system for CoMeDi Shared Task, focusing on Subtask 2: Disagreement Ranking. Our system leverages sentence embeddings generated by the paraphrase-xlm-r-multilingual-v1 model, combined with a deep neural regression model incorporating batch normalization and dropout for improved generalization. By predicting the mean of pairwise judgment differences between annotators, our method explicitly targets disagreement ranking, diverging from traditional "gold label" aggregation approaches. We optimized our system with a customized architecture and training procedure, achieving competitive performance in Spearman correlation against mean disagreement labels. Our results highlight the importance of robust embeddings, effective model architecture, and careful handling of judgment differences for ranking disagreement in multilingual contexts. These findings provide insights into the use of contextualized representations for ordinal judgment tasks and open avenues for further refinement of disagreement prediction models.
2501.12337
Understanding User Preference -- Comparison between Linear and Directional Top-K Query results
cs.DB
This paper investigates user preferences for Linear Top-k Queries and Directional Top-k Queries, two methods for ranking results in multidimensional datasets. While Linear Queries prioritize weighted sums of attributes, Directional Queries aim to deliver more balanced results by incorporating the spatial relationship between data points and a user-defined preference line. The study explores how preferences for these methods vary across different contexts by focusing on two real-world topics: used cars (e-commerce domain) and football players (personal interest domain). A user survey involving 106 participants was conducted to evaluate preferences, with results visualized as scatter plots for comparison. The findings reveal a significant preference for directional queries in the used cars topic, where balanced results align better with user goals. In contrast, preferences in the football players topic were more evenly distributed, influenced by user expertise and familiarity with the domain. Additionally, the study demonstrates that the two specific topics selected for this research exhibit significant differences in their impact on user preferences. This research reveals authentic user preferences, highlighting the practical utility of Directional Queries for lifestyle-related applications and the subjective nature of preferences in specialized domains. These insights contribute to advancing personalized database technologies, guiding the development of more user-centric ranking systems.
2501.12339
Treefix: Enabling Execution with a Tree of Prefixes
cs.SE cs.AI
The ability to execute code is a prerequisite for various dynamic program analyses. Learning-guided execution has been proposed as an approach to enable the execution of arbitrary code snippets by letting a neural model predict likely values for any missing variables. Although state-of-the-art learning-guided execution approaches, such as LExecutor, can enable the execution of a relative high amount of code, they are limited to predicting a restricted set of possible values and do not use any feedback from previous executions to execute even more code. This paper presents Treefix, a novel learning-guided execution approach that leverages LLMs to iteratively create code prefixes that enable the execution of a given code snippet. The approach addresses the problem in a multi-step fashion, where each step uses feedback about the code snippet and its execution to instruct an LLM to improve a previously generated prefix. This process iteratively creates a tree of prefixes, a subset of which is returned to the user as prefixes that maximize the number of executed lines in the code snippet. In our experiments with two datasets of Python code snippets, Treefix achieves 25% and 7% more coverage relative to the current state of the art in learning-guided execution, covering a total of 84% and 82% of all lines in the code snippets.
2501.12344
CYCle: Choosing Your Collaborators Wisely to Enhance Collaborative Fairness in Decentralized Learning
cs.LG cs.DC
Collaborative learning (CL) enables multiple participants to jointly train machine learning (ML) models on decentralized data sources without raw data sharing. While the primary goal of CL is to maximize the expected accuracy gain for each participant, it is also important to ensure that the gains are fairly distributed. Specifically, no client should be negatively impacted by the collaboration, and the individual gains must ideally be commensurate with the contributions. Most existing CL algorithms require central coordination and focus on the gain maximization objective while ignoring collaborative fairness. In this work, we first show that the existing measure of collaborative fairness based on the correlation between accuracy values without and with collaboration has drawbacks because it does not account for negative collaboration gain. We argue that maximizing mean collaboration gain (MCG) while simultaneously minimizing the collaboration gain spread (CGS) is a fairer alternative. Next, we propose the CYCle protocol that enables individual participants in a private decentralized learning (PDL) framework to achieve this objective through a novel reputation scoring method based on gradient alignment between the local cross-entropy and distillation losses. Experiments on the CIFAR-10, CIFAR-100, and Fed-ISIC2019 datasets empirically demonstrate the effectiveness of the CYCle protocol to ensure positive and fair collaboration gain for all participants, even in cases where the data distributions of participants are highly skewed. For the simple mean estimation problem with two participants, we also theoretically show that CYCle performs better than standard FedAvg, especially when there is large statistical heterogeneity.
2501.12348
Rate-Distortion-Perception Function of Bernoulli Vector Sources
cs.IT math.IT
In this paper, we consider the rate-distortion-perception (RDP) trade-off for the lossy compression of a Bernoulli vector source, which is a finite collection of independent binary random variables. The RDP function quantifies in a way the efficient compression of a source when we impose a distortion constraint that limits the dissimilarity between the source and the reconstruction and a perception constraint that restricts the distributional discrepancy of the source and the reconstruction. In this work, we obtain an exact characterization of the RDP function of a Bernoulli vector source with the Hamming distortion function and a single-letter perception function that measures the closeness of the distributions of the components of the source. The solution can be described by partitioning the set of distortion and perception levels $(D,P)$ into three regions, where in each region the optimal distortion and perception levels we allot to the components have a similar nature. Finally, we introduce the RDP function for graph sources and apply our result to the Erd\H{o}s-R\'enyi graph model.
2501.12349
General Field Evaluation in High-Order Meshes on GPUs
cs.MS cs.CE
Robust and scalable function evaluation at any arbitrary point in the finite/spectral element mesh is required for querying the partial differential equation solution at points of interest, comparison of solution between different meshes, and Lagrangian particle tracking. This is a challenging problem, particularly for high-order unstructured meshes partitioned in parallel with MPI, as it requires identifying the element that overlaps a given point and computing the corresponding reference space coordinates. We present a robust and efficient technique for general field evaluation in large-scale high-order meshes with quadrilaterals and hexahedra. In the proposed method, a combination of globally partitioned and processor-local maps are used to first determine a list of candidate MPI ranks, and then locally candidate elements that could contain a given point. Next, element-wise bounding boxes further reduce the list of candidate elements. Finally, Newton's method with trust region is used to determine the overlapping element and corresponding reference space coordinates. Since GPU-based architectures have become popular for accelerating computational analyses using meshes with tensor-product elements, specialized kernels have been developed to utilize the proposed methodology on GPUs. The method is also extended to enable general field evaluation on surface meshes. The paper concludes by demonstrating the use of proposed method in various applications ranging from mesh-to-mesh transfer during r-adaptivity to Lagrangian particle tracking.
2501.12352
Test-time regression: a unifying framework for designing sequence models with associative memory
cs.LG cs.AI cs.NE stat.ML
Sequences provide a remarkably general way to represent and process information. This powerful abstraction has placed sequence modeling at the center of modern deep learning applications, inspiring numerous architectures from transformers to recurrent networks. While this fragmented development has yielded powerful models, it has left us without a unified framework to understand their fundamental similarities and explain their effectiveness. We present a unifying framework motivated by an empirical observation: effective sequence models must be able to perform associative recall. Our key insight is that memorizing input tokens through an associative memory is equivalent to performing regression at test-time. This regression-memory correspondence provides a framework for deriving sequence models that can perform associative recall, offering a systematic lens to understand seemingly ad-hoc architectural choices. We show numerous recent architectures -- including linear attention models, their gated variants, state-space models, online learners, and softmax attention -- emerge naturally as specific approaches to test-time regression. Each architecture corresponds to three design choices: the relative importance of each association, the regressor function class, and the optimization algorithm. This connection leads to new understanding: we provide theoretical justification for QKNorm in softmax attention, and we motivate higher-order generalizations of softmax attention. Beyond unification, our work unlocks decades of rich statistical tools that can guide future development of more powerful yet principled sequence models.
2501.12354
Diffusion-aware Censored Gaussian Processes for Demand Modelling
cs.LG stat.ME stat.ML
Inferring the true demand for a product or a service from aggregate data is often challenging due to the limited available supply, thus resulting in observations that are censored and correspond to the realized demand, thereby not accounting for the unsatisfied demand. Censored regression models are able to account for the effect of censoring due to the limited supply, but they don't consider the effect of substitutions, which may cause the demand for similar alternative products or services to increase. This paper proposes Diffusion-aware Censored Demand Models, which combine a Tobit likelihood with a graph diffusion process in order to model the latent process of transfer of unsatisfied demand between similar products or services. We instantiate this new class of models under the framework of GPs and, based on both simulated and real-world data for modeling sales, bike-sharing demand, and EV charging demand, demonstrate its ability to better recover the true demand and produce more accurate out-of-sample predictions.
2501.12356
Vision-Language Models for Automated Chest X-ray Interpretation: Leveraging ViT and GPT-2
cs.CV
Radiology plays a pivotal role in modern medicine due to its non-invasive diagnostic capabilities. However, the manual generation of unstructured medical reports is time consuming and prone to errors. It creates a significant bottleneck in clinical workflows. Despite advancements in AI-generated radiology reports, challenges remain in achieving detailed and accurate report generation. In this study we have evaluated different combinations of multimodal models that integrate Computer Vision and Natural Language Processing to generate comprehensive radiology reports. We employed a pretrained Vision Transformer (ViT-B16) and a SWIN Transformer as the image encoders. The BART and GPT-2 models serve as the textual decoders. We used Chest X-ray images and reports from the IU-Xray dataset to evaluate the usability of the SWIN Transformer-BART, SWIN Transformer-GPT-2, ViT-B16-BART and ViT-B16-GPT-2 models for report generation. We aimed at finding the best combination among the models. The SWIN-BART model performs as the best-performing model among the four models achieving remarkable results in almost all the evaluation metrics like ROUGE, BLEU and BERTScore.
2501.12359
Measured Hockey-Stick Divergence and its Applications to Quantum Pufferfish Privacy
quant-ph cs.CR cs.IT cs.LG math.IT
The hockey-stick divergence is a fundamental quantity characterizing several statistical privacy frameworks that ensure privacy for classical and quantum data. In such quantum privacy frameworks, the adversary is allowed to perform all possible measurements. However, in practice, there are typically limitations to the set of measurements that can be performed. To this end, here, we comprehensively analyze the measured hockey-stick divergence under several classes of practically relevant measurement classes. We prove several of its properties, including data processing and convexity. We show that it is efficiently computable by semi-definite programming for some classes of measurements and can be analytically evaluated for Werner and isotropic states. Notably, we show that the measured hockey-stick divergence characterizes optimal privacy parameters in the quantum pufferfish privacy framework. With this connection and the developed technical tools, we enable methods to quantify and audit privacy for several practically relevant settings. Lastly, we introduce the measured hockey-stick divergence of channels and explore its applications in ensuring privacy for channels.
2501.12362
ARM-IRL: Adaptive Resilience Metric Quantification Using Inverse Reinforcement Learning
eess.SY cs.SY
Resilience of safety-critical systems is gaining importance, particularly with the increasing number of cyber and physical threats. Cyber-physical threats are becoming increasingly prevalent, as digital systems are ubiquitous in critical infrastructure. The challenge with determining the resilience of cyber-physical systems is identifying a set of resilience metrics that can adapt to the changing states of the system. A static resilience metric can lead to an inaccurate estimation of system state, and can result in unintended consequences against cyber threats. In this work, we propose a data-driven method for adaptive resilience metric learning. The primary goal is to learn a single resilience metric by formulating an inverse reinforcement learning problem that learns a reward or objective from a set of control actions from an expert. It learns the structure or parameters of the reward function based on information provided by expert demonstrations. Most prior work has considered static weights or theories from fuzzy logic to formulate a single resilience metric. Instead, this work learns the resilience metric, represented as reward function, using adversarial inverse reinforcement learning, to determine the optimal policy through training the generator discriminator in parallel. We evaluate our proposed technique in scenarios such as optimal communication network rerouting, power distribution network reconfiguration, and a combined cyber-physical restoration of critical load using the IEEE 123-bus system.
2501.12365
Efficient Algorithm for Sparse Fourier Transform of Generalized q-ary Functions
cs.CC cs.DM cs.IT cs.LG math.IT
Computing the Fourier transform of a $q$-ary function $f:\mathbb{Z}_{q}^n\rightarrow \mathbb{R}$, which maps $q$-ary sequences to real numbers, is an important problem in mathematics with wide-ranging applications in biology, signal processing, and machine learning. Previous studies have shown that, under the sparsity assumption, the Fourier transform can be computed efficiently using fast and sample-efficient algorithms. However, in many practical settings, the function is defined over a more general space -- the space of generalized $q$-ary sequences $\mathbb{Z}_{q_1} \times \mathbb{Z}_{q_2} \times \cdots \times \mathbb{Z}_{q_n}$ -- where each $\mathbb{Z}_{q_i}$ corresponds to integers modulo $q_i$. A naive approach involves setting $q=\max_i{q_i}$ and treating the function as $q$-ary, which results in heavy computational overheads. Herein, we develop GFast, an algorithm that computes the $S$-sparse Fourier transform of $f$ with a sample complexity of $O(Sn)$, computational complexity of $O(Sn \log N)$, and a failure probability that approaches zero as $N=\prod_{i=1}^n q_i \rightarrow \infty$ with $S = N^\delta$ for some $0 \leq \delta < 1$. In the presence of noise, we further demonstrate that a robust version of GFast computes the transform with a sample complexity of $O(Sn^2)$ and computational complexity of $O(Sn^2 \log N)$ under the same high probability guarantees. Using large-scale synthetic experiments, we demonstrate that GFast computes the sparse Fourier transform of generalized $q$-ary functions using $16\times$ fewer samples and running $8\times$ faster than existing algorithms. In real-world protein fitness datasets, GFast explains the predictive interactions of a neural network with $>25\%$ smaller normalized mean-squared error compared to existing algorithms.
2501.12367
Budget-constrained Collaborative Renewable Energy Forecasting Market
cs.LG
Accurate power forecasting from renewable energy sources (RES) is crucial for integrating additional RES capacity into the power system and realizing sustainability goals. This work emphasizes the importance of integrating decentralized spatio-temporal data into forecasting models. However, decentralized data ownership presents a critical obstacle to the success of such spatio-temporal models, and incentive mechanisms to foster data-sharing need to be considered. The main contributions are a) a comparative analysis of the forecasting models, advocating for efficient and interpretable spline LASSO regression models, and b) a bidding mechanism within the data/analytics market to ensure fair compensation for data providers and enable both buyers and sellers to express their data price requirements. Furthermore, an incentive mechanism for time series forecasting is proposed, effectively incorporating price constraints and preventing redundant feature allocation. Results show significant accuracy improvements and potential monetary gains for data sellers. For wind power data, an average root mean squared error improvement of over 10% was achieved by comparing forecasts generated by the proposal with locally generated ones.
2501.12368
InternLM-XComposer2.5-Reward: A Simple Yet Effective Multi-Modal Reward Model
cs.CV cs.CL
Despite the promising performance of Large Vision Language Models (LVLMs) in visual understanding, they occasionally generate incorrect outputs. While reward models (RMs) with reinforcement learning or test-time scaling offer the potential for improving generation quality, a critical gap remains: publicly available multi-modal RMs for LVLMs are scarce, and the implementation details of proprietary models are often unclear. We bridge this gap with InternLM-XComposer2.5-Reward (IXC-2.5-Reward), a simple yet effective multi-modal reward model that aligns LVLMs with human preferences. To ensure the robustness and versatility of IXC-2.5-Reward, we set up a high-quality multi-modal preference corpus spanning text, image, and video inputs across diverse domains, such as instruction following, general understanding, text-rich documents, mathematical reasoning, and video understanding. IXC-2.5-Reward achieves excellent results on the latest multi-modal reward model benchmark and shows competitive performance on text-only reward model benchmarks. We further demonstrate three key applications of IXC-2.5-Reward: (1) Providing a supervisory signal for RL training. We integrate IXC-2.5-Reward with Proximal Policy Optimization (PPO) yields IXC-2.5-Chat, which shows consistent improvements in instruction following and multi-modal open-ended dialogue; (2) Selecting the best response from candidate responses for test-time scaling; and (3) Filtering outlier or noisy samples from existing image and video instruction tuning training data. To ensure reproducibility and facilitate further research, we have open-sourced all model weights and training recipes at https://github.com/InternLM/InternLM-XComposer
2501.12369
DARB-Splatting: Generalizing Splatting with Decaying Anisotropic Radial Basis Functions
cs.CV cs.AI cs.GR
Splatting-based 3D reconstruction methods have gained popularity with the advent of 3D Gaussian Splatting, efficiently synthesizing high-quality novel views. These methods commonly resort to using exponential family functions, such as the Gaussian function, as reconstruction kernels due to their anisotropic nature, ease of projection, and differentiability in rasterization. However, the field remains restricted to variations within the exponential family, leaving generalized reconstruction kernels largely underexplored, partly due to the lack of easy integrability in 3D to 2D projections. In this light, we show that a class of decaying anisotropic radial basis functions (DARBFs), which are non-negative functions of the Mahalanobis distance, supports splatting by approximating the Gaussian function's closed-form integration advantage. With this fresh perspective, we demonstrate up to 34% faster convergence during training and a 15% reduction in memory consumption across various DARB reconstruction kernels, while maintaining comparable PSNR, SSIM, and LPIPS results. We will make the code available.
2501.12370
Parameters vs FLOPs: Scaling Laws for Optimal Sparsity for Mixture-of-Experts Language Models
cs.LG cs.AI
Scaling the capacity of language models has consistently proven to be a reliable approach for improving performance and unlocking new capabilities. Capacity can be primarily defined by two dimensions: the number of model parameters and the compute per example. While scaling typically involves increasing both, the precise interplay between these factors and their combined contribution to overall capacity remains not fully understood. We explore this relationship in the context of sparse Mixture-of-Experts (MoEs), which allow scaling the number of parameters without proportionally increasing the FLOPs per example. We investigate how varying the sparsity level, i.e., the fraction of inactive parameters, impacts model's performance during pretraining and downstream few-shot evaluation. We find that under different constraints (e.g., parameter size and total training compute), there is an optimal level of sparsity that improves both training efficiency and model performance. These results provide a better understanding of the impact of sparsity in scaling laws for MoEs and complement existing works in this area, offering insights for designing more efficient architectures.
2501.12371
CAT and DOG: Improved Codes for Private Distributed Matrix Multiplication
cs.IT math.IT
We present novel constructions of polynomial codes for private distributed matrix multiplication (PDMM/SDMM) using outer product partitioning (OPP). We extend the degree table framework from the literature to cyclic-addition degree tables (CATs). By using roots of unity as evaluation points, we enable modulo-addition in the table. Based on CATs, we present an explicit construction, called CATx, that requires fewer workers than existing schemes in the low-privacy regime. Additionally, we present new families of schemes based on conventional degree tables, called GASPrs and DOGrs, that outperform the state-of-the-art for a wide range of parameters.
2501.12372
Is Long Context All You Need? Leveraging LLM's Extended Context for NL2SQL
cs.DB cs.AI
Large Language Models (LLMs) have demonstrated impressive capabilities across a range of natural language processing tasks. In particular, improvements in reasoning abilities and the expansion of context windows have opened new avenues for leveraging these powerful models. NL2SQL is challenging in that the natural language question is inherently ambiguous, while the SQL generation requires a precise understanding of complex data schema and semantics. One approach to this semantic ambiguous problem is to provide more and sufficient contextual information. In this work, we explore the performance and the latency trade-offs of the extended context window (a.k.a., long context) offered by Google's state-of-the-art LLM (\textit{gemini-1.5-pro}). We study the impact of various contextual information, including column example values, question and SQL query pairs, user-provided hints, SQL documentation, and schema. To the best of our knowledge, this is the first work to study how the extended context window and extra contextual information can help NL2SQL generation with respect to both accuracy and latency cost. We show that long context LLMs are robust and do not get lost in the extended contextual information. Additionally, our long-context NL2SQL pipeline based on Google's \textit{gemini-pro-1.5} achieve strong performances on various benchmark datasets without finetuning and expensive self-consistency based techniques.
2501.12374
Expertise elevates AI usage: experimental evidence comparing laypeople and professional artists
cs.HC cs.AI cs.CY
Novel capacities of generative AI to analyze and generate cultural artifacts raise inevitable questions about the nature and value of artistic education and human expertise. Has AI already leveled the playing field between professional artists and laypeople, or do trained artistic expressive capacity, curation skills and experience instead enhance the ability to use these new tools? In this pre-registered study, we conduct experimental comparisons between 50 active artists and a demographically matched sample of laypeople. We designed two tasks to approximate artistic practice for testing their capabilities in both faithful and creative image creation: replicating a reference image, and moving as far away as possible from it. We developed a bespoke platform where participants used a modern text-to-image model to complete both tasks. We also collected and compared participants' sentiments towards AI. On average, artists produced more faithful and creative outputs than their lay counterparts, although only by a small margin. While AI may ease content creation, professional expertise is still valuable - even within the confined space of generative AI itself. Finally, we also explored how well an exemplary vision-capable large language model (GPT-4o) would complete the same tasks, if given the role of an image generation agent, and found it performed on par in copying but outperformed even artists in the creative task. The very best results were still produced by humans in both tasks. These outcomes highlight the importance of integrating artistic skills with AI training to prepare artists and other visual professionals for a technologically evolving landscape. We see a potential in collaborative synergy with generative AI, which could reshape creative industries and education in the arts.
2501.12375
Video Depth Anything: Consistent Depth Estimation for Super-Long Videos
cs.CV cs.AI
Depth Anything has achieved remarkable success in monocular depth estimation with strong generalization ability. However, it suffers from temporal inconsistency in videos, hindering its practical applications. Various methods have been proposed to alleviate this issue by leveraging video generation models or introducing priors from optical flow and camera poses. Nonetheless, these methods are only applicable to short videos (< 10 seconds) and require a trade-off between quality and computational efficiency. We propose Video Depth Anything for high-quality, consistent depth estimation in super-long videos (over several minutes) without sacrificing efficiency. We base our model on Depth Anything V2 and replace its head with an efficient spatial-temporal head. We design a straightforward yet effective temporal consistency loss by constraining the temporal depth gradient, eliminating the need for additional geometric priors. The model is trained on a joint dataset of video depth and unlabeled images, similar to Depth Anything V2. Moreover, a novel key-frame-based strategy is developed for long video inference. Experiments show that our model can be applied to arbitrarily long videos without compromising quality, consistency, or generalization ability. Comprehensive evaluations on multiple video benchmarks demonstrate that our approach sets a new state-of-the-art in zero-shot video depth estimation. We offer models of different scales to support a range of scenarios, with our smallest model capable of real-time performance at 30 FPS.
2501.12379
Constant Weight Polar Codes through Periodic Markov Processes
cs.IT math.IT
Constant weight codes can arise from an input process sampled from a periodic Markov chain. A previous result showed that, in general, polarization does not occur for input-output processes with an underlying periodic Markov chain. In this work, we show that if we fix the initial state of an underlying periodic Markov chain, polarization does occur. Fixing the initial state is aligned with ensuring a constant weight code.
2501.12380
MMVU: Measuring Expert-Level Multi-Discipline Video Understanding
cs.CV cs.AI cs.CL
We introduce MMVU, a comprehensive expert-level, multi-discipline benchmark for evaluating foundation models in video understanding. MMVU includes 3,000 expert-annotated questions spanning 27 subjects across four core disciplines: Science, Healthcare, Humanities & Social Sciences, and Engineering. Compared to prior benchmarks, MMVU features three key advancements. First, it challenges models to apply domain-specific knowledge and perform expert-level reasoning to analyze specialized-domain videos, moving beyond the basic visual perception typically assessed in current video benchmarks. Second, each example is annotated by human experts from scratch. We implement strict data quality controls to ensure the high quality of the dataset. Finally, each example is enriched with expert-annotated reasoning rationals and relevant domain knowledge, facilitating in-depth analysis. We conduct an extensive evaluation of 32 frontier multimodal foundation models on MMVU. The latest System-2-capable models, o1 and Gemini 2.0 Flash Thinking, achieve the highest performance among the tested models. However, they still fall short of matching human expertise. Through in-depth error analyses and case studies, we offer actionable insights for future advancements in expert-level, knowledge-intensive video understanding for specialized domains.
2501.12381
Parallel Sequence Modeling via Generalized Spatial Propagation Network
cs.CV cs.LG
We present the Generalized Spatial Propagation Network (GSPN), a new attention mechanism optimized for vision tasks that inherently captures 2D spatial structures. Existing attention models, including transformers, linear attention, and state-space models like Mamba, process multi-dimensional data as 1D sequences, compromising spatial coherence and efficiency. GSPN overcomes these limitations by directly operating on spatially coherent image data and forming dense pairwise connections through a line-scan approach. Central to GSPN is the Stability-Context Condition, which ensures stable, context-aware propagation across 2D sequences and reduces the effective sequence length to $\sqrt{N}$ for a square map with N elements, significantly enhancing computational efficiency. With learnable, input-dependent weights and no reliance on positional embeddings, GSPN achieves superior spatial fidelity and state-of-the-art performance in vision tasks, including ImageNet classification, class-guided image generation, and text-to-image generation. Notably, GSPN accelerates SD-XL with softmax-attention by over $84\times$ when generating 16K images.
2501.12382
DiffDoctor: Diagnosing Image Diffusion Models Before Treating
cs.CV
In spite of the recent progress, image diffusion models still produce artifacts. A common solution is to refine an established model with a quality assessment system, which generally rates an image in its entirety. In this work, we believe problem-solving starts with identification, yielding the request that the model should be aware of not just the presence of defects in an image, but their specific locations. Motivated by this, we propose DiffDoctor, a two-stage pipeline to assist image diffusion models in generating fewer artifacts. Concretely, the first stage targets developing a robust artifact detector, for which we collect a dataset of over 1M flawed synthesized images and set up an efficient human-in-the-loop annotation process, incorporating a carefully designed class-balance strategy. The learned artifact detector is then involved in the second stage to tune the diffusion model through assigning a per-pixel confidence map for each synthesis. Extensive experiments on text-to-image diffusion models demonstrate the effectiveness of our artifact detector as well as the soundness of our diagnose-then-treat design.
2501.12384
CCESAR: Coastline Classification-Extraction From SAR Images Using CNN-U-Net Combination
cs.CV cs.LG eess.IV
In this article, we improve the deep learning solution for coastline extraction from Synthetic Aperture Radar (SAR) images by proposing a two-stage model involving image classification followed by segmentation. We hypothesize that a single segmentation model usually used for coastline detection is insufficient to characterize different coastline types. We demonstrate that the need for a two-stage workflow prevails through different compression levels of these images. Our results from experiments using a combination of CNN and U-Net models on Sentinel-1 images show that the two-stage workflow, coastline classification-extraction from SAR images (CCESAR) outperforms a single U-Net segmentation model.
2501.12385
Audio Texture Manipulation by Exemplar-Based Analogy
cs.SD cs.LG eess.AS
Audio texture manipulation involves modifying the perceptual characteristics of a sound to achieve specific transformations, such as adding, removing, or replacing auditory elements. In this paper, we propose an exemplar-based analogy model for audio texture manipulation. Instead of conditioning on text-based instructions, our method uses paired speech examples, where one clip represents the original sound and another illustrates the desired transformation. The model learns to apply the same transformation to new input, allowing for the manipulation of sound textures. We construct a quadruplet dataset representing various editing tasks, and train a latent diffusion model in a self-supervised manner. We show through quantitative evaluations and perceptual studies that our model outperforms text-conditioned baselines and generalizes to real-world, out-of-distribution, and non-speech scenarios. Project page: https://berkeley-speech-group.github.io/audio-texture-analogy/
2501.12386
InternVideo2.5: Empowering Video MLLMs with Long and Rich Context Modeling
cs.CV
This paper aims to improve the performance of video multimodal large language models (MLLM) via long and rich context (LRC) modeling. As a result, we develop a new version of InternVideo2.5 with a focus on enhancing the original MLLMs' ability to perceive fine-grained details and capture long-form temporal structure in videos. Specifically, our approach incorporates dense vision task annotations into MLLMs using direct preference optimization and develops compact spatiotemporal representations through adaptive hierarchical token compression. Experimental results demonstrate this unique design of LRC greatly improves the results of video MLLM in mainstream video understanding benchmarks (short & long), enabling the MLLM to memorize significantly longer video inputs (at least 6x longer than the original), and master specialized vision capabilities like object tracking and segmentation. Our work highlights the importance of multimodal context richness (length and fineness) in empowering MLLM's innate abilites (focus and memory), providing new insights for future research on video MLLM. Code and models are available at https://github.com/OpenGVLab/InternVideo/tree/main/InternVideo2.5
2501.12387
Continuous 3D Perception Model with Persistent State
cs.CV
We present a unified framework capable of solving a broad range of 3D tasks. Our approach features a stateful recurrent model that continuously updates its state representation with each new observation. Given a stream of images, this evolving state can be used to generate metric-scale pointmaps (per-pixel 3D points) for each new input in an online fashion. These pointmaps reside within a common coordinate system, and can be accumulated into a coherent, dense scene reconstruction that updates as new images arrive. Our model, called CUT3R (Continuous Updating Transformer for 3D Reconstruction), captures rich priors of real-world scenes: not only can it predict accurate pointmaps from image observations, but it can also infer unseen regions of the scene by probing at virtual, unobserved views. Our method is simple yet highly flexible, naturally accepting varying lengths of images that may be either video streams or unordered photo collections, containing both static and dynamic content. We evaluate our method on various 3D/4D tasks and demonstrate competitive or state-of-the-art performance in each. Project Page: https://cut3r.github.io/
2501.12389
Taming Teacher Forcing for Masked Autoregressive Video Generation
cs.CV
We introduce MAGI, a hybrid video generation framework that combines masked modeling for intra-frame generation with causal modeling for next-frame generation. Our key innovation, Complete Teacher Forcing (CTF), conditions masked frames on complete observation frames rather than masked ones (namely Masked Teacher Forcing, MTF), enabling a smooth transition from token-level (patch-level) to frame-level autoregressive generation. CTF significantly outperforms MTF, achieving a +23% improvement in FVD scores on first-frame conditioned video prediction. To address issues like exposure bias, we employ targeted training strategies, setting a new benchmark in autoregressive video generation. Experiments show that MAGI can generate long, coherent video sequences exceeding 100 frames, even when trained on as few as 16 frames, highlighting its potential for scalable, high-quality video generation.
2501.12390
GPS as a Control Signal for Image Generation
cs.CV
We show that the GPS tags contained in photo metadata provide a useful control signal for image generation. We train GPS-to-image models and use them for tasks that require a fine-grained understanding of how images vary within a city. In particular, we train a diffusion model to generate images conditioned on both GPS and text. The learned model generates images that capture the distinctive appearance of different neighborhoods, parks, and landmarks. We also extract 3D models from 2D GPS-to-image models through score distillation sampling, using GPS conditioning to constrain the appearance of the reconstruction from each viewpoint. Our evaluations suggest that our GPS-conditioned models successfully learn to generate images that vary based on location, and that GPS conditioning improves estimated 3D structure.
2501.12391
Physics of Skill Learning
cs.LG cs.AI physics.data-an stat.ML
We aim to understand physics of skill learning, i.e., how skills are learned in neural networks during training. We start by observing the Domino effect, i.e., skills are learned sequentially, and notably, some skills kick off learning right after others complete learning, similar to the sequential fall of domino cards. To understand the Domino effect and relevant behaviors of skill learning, we take physicists' approach of abstraction and simplification. We propose three models with varying complexities -- the Geometry model, the Resource model, and the Domino model, trading between reality and simplicity. The Domino effect can be reproduced in the Geometry model, whose resource interpretation inspires the Resource model, which can be further simplified to the Domino model. These models present different levels of abstraction and simplification; each is useful to study some aspects of skill learning. The Geometry model provides interesting insights into neural scaling laws and optimizers; the Resource model sheds light on the learning dynamics of compositional tasks; the Domino model reveals the benefits of modularity. These models are not only conceptually interesting -- e.g., we show how Chinchilla scaling laws can emerge from the Geometry model, but also are useful in practice by inspiring algorithmic development -- e.g., we show how simple algorithmic changes, motivated by these toy models, can speed up the training of deep learning models.
2501.12392
Learning segmentation from point trajectories
cs.CV cs.AI cs.LG
We consider the problem of segmenting objects in videos based on their motion and no other forms of supervision. Prior work has often approached this problem by using the principle of common fate, namely the fact that the motion of points that belong to the same object is strongly correlated. However, most authors have only considered instantaneous motion from optical flow. In this work, we present a way to train a segmentation network using long-term point trajectories as a supervisory signal to complement optical flow. The key difficulty is that long-term motion, unlike instantaneous motion, is difficult to model -- any parametric approximation is unlikely to capture complex motion patterns over long periods of time. We instead draw inspiration from subspace clustering approaches, proposing a loss function that seeks to group the trajectories into low-rank matrices where the motion of object points can be approximately explained as a linear combination of other point tracks. Our method outperforms the prior art on motion-based segmentation, which shows the utility of long-term motion and the effectiveness of our formulation.
2501.12393
Towards Affordance-Aware Articulation Synthesis for Rigged Objects
cs.CV
Rigged objects are commonly used in artist pipelines, as they can flexibly adapt to different scenes and postures. However, articulating the rigs into realistic affordance-aware postures (e.g., following the context, respecting the physics and the personalities of the object) remains time-consuming and heavily relies on human labor from experienced artists. In this paper, we tackle the novel problem and design A3Syn. With a given context, such as the environment mesh and a text prompt of the desired posture, A3Syn synthesizes articulation parameters for arbitrary and open-domain rigged objects obtained from the Internet. The task is incredibly challenging due to the lack of training data, and we do not make any topological assumptions about the open-domain rigs. We propose using 2D inpainting diffusion model and several control techniques to synthesize in-context affordance information. Then, we develop an efficient bone correspondence alignment using a combination of differentiable rendering and semantic correspondence. A3Syn has stable convergence, completes in minutes, and synthesizes plausible affordance on different combinations of in-the-wild object rigs and scenes.
2501.12394
The ELEVATE-AI LLMs Framework: An Evaluation Framework for Use of Large Language Models in HEOR: an ISPOR Working Group Report
cs.CY cs.LG
Introduction. Generative Artificial Intelligence, particularly large language models (LLMs), offers transformative potential for Health Economics and Outcomes Research (HEOR). However, evaluating the quality, transparency, and rigor of LLM-assisted research lacks standardized guidance. This article introduces the ELEVATE AI LLMs framework and checklist, designed to support researchers and reviewers in assessing LLM use in HEOR. Methods. The ELEVATE AI LLMs framework was developed through a targeted review of existing guidelines and evaluation frameworks. The framework comprises ten evaluation domains, including model characteristics, accuracy, comprehensiveness, and fairness. The accompanying checklist operationalizes the framework. To validate the framework, we applied it to two published studies, demonstrating its usability across different HEOR tasks. Results. The ELEVATE AI LLMs framework provides a comprehensive structure for evaluating LLM-assisted research, while the checklist facilitates practical application. Validation of the framework and checklist on studies of systematic literature reviews and health economic modeling highlighted their ability to identify strengths and gaps in reporting. Limitations. While the ELEVATE AI LLMs framework provides robust guidance, its broader generalizability and applicability to diverse HEOR tasks require further empirical testing. Additionally, several metrics adapted from computer science need further validation in HEOR contexts. Conclusion. The ELEVATE AI LLMs framework and checklist fill a critical gap in HEOR by offering structured guidance for evaluating LLM-assisted research. By promoting transparency, accuracy, and reproducibility, they aim to standardize and improve the integration of LLMs into HEOR, ensuring their outputs meet the field's rigorous standards.
2501.12399
FinSphere: A Conversational Stock Analysis Agent Equipped with Quantitative Tools based on Real-Time Database
cs.AI cs.CL cs.IR q-fin.CP
Current financial Large Language Models (LLMs) struggle with two critical limitations: a lack of depth in stock analysis, which impedes their ability to generate professional-grade insights, and the absence of objective evaluation metrics to assess the quality of stock analysis reports. To address these challenges, this paper introduces FinSphere, a conversational stock analysis agent, along with three major contributions: (1) Stocksis, a dataset curated by industry experts to enhance LLMs' stock analysis capabilities, (2) AnalyScore, a systematic evaluation framework for assessing stock analysis quality, and (3) FinSphere, an AI agent that can generate high-quality stock analysis reports in response to user queries. Experiments demonstrate that FinSphere achieves superior performance compared to both general and domain-specific LLMs, as well as existing agent-based systems, even when they are enhanced with real-time data access and few-shot guidance. The integrated framework, which combines real-time data feeds, quantitative tools, and an instruction-tuned LLM, yields substantial improvements in both analytical quality and practical applicability for real-world stock analysis.