arXiv subjects

Xiaofang Wang

Publications and source records attributed to Xiaofang Wang.

At least 19 recordsLinked to original sources

arXiv2026

Information-Regularized Attention for Visual-Centric Reasoning

Vision-language models (VLMs) have become a paradigm for multimodal learning, yet remain unstable due to object hallucination, weak visual grounding, and catastrophic forgetting after full-parameter instruction tuning. We claim these failures result from a lack of explicit control over visual representation learning during the standard next-token prediction objective. As a result, visual embeddings thus become passively optimized and prone to injecting redundant or spurious signals. To counter this, we introduce Information-Regularized Attention (IRA), a stochastic attention mechanism that explicitly regulates the amount of visual information injected into the hidden states of intermediate transformer layers. This local reparameterization translates uncertainty about visual representations into local noise that is independent across data points. Beyond evaluating model performance, we also quantify embedding properties, where IRA produces smoother curvature trajectories and suppresses attention-sink across all layers, indicating a more stable transformation of the visual signal. Our results suggest that stochastic attention is not merely a regularizer but a key contributor to representation learning in a generative architecture, offering a new direction for building more reliable VLMs.

cs.CV↗

arXiv2026

Building a Mind Palace: Structuring Environment-Grounded Semantic Graphs for Effective Long Video Analysis with LLMs

Long-form video understanding with Large Vision Language Models is challenged by the need to analyze temporally dispersed yet spatially concentrated key moments within limited context windows. In this work, we introduce VideoMindPalace, a new framework inspired by the "Mind Palace", which organizes critical video moments into a topologically structured semantic graph. VideoMindPalace organizes key information through (i) hand-object tracking and interaction, (ii) clustered activity zones representing specific areas of recurring activities, and (iii) environment layout mapping, allowing natural language parsing by LLMs to provide grounded insights on spatio-temporal and 3D context. In addition, we propose the Video MindPalace Benchmark (VMB), to assess human-like reasoning, including spatial localization, temporal reasoning, and layout-aware sequential understanding. Evaluated on VMB and established video QA datasets, including EgoSchema, NExT-QA, IntentQA, and the Active Memories Benchmark, VideoMindPalace demonstrates notable gains in spatio-temporal coherence and human-aligned reasoning, advancing long-form video analysis capabilities in VLMs.

cs.CV↗

arXiv2025

Soliton eigenvalue control by interaction of circularly polarized lights in a nonlinear fiber

We propose a physical method for controlling soliton eigenvalues in optical fibers, which is realized through the interaction between circularly polarized lights. Using this method, we not only achieve the decomposition of high-order solitons (HOSs) with different orders, but also realize physical processes of reconstructing HOSs for the first time. Compared with existing methods, our approach ensures accurate measurement of the discrete eigenvalues of HOSs while exhibiting higher decomposition efficiency. It is worth noting that the probe soliton, which induces these phenomena, plays a key role. The requirement for a moderate steepness of the probe suggests the presence of an uncertainty principle in the measurement of soliton eigenvalues, similar to the detection of microscopic particles. Our results can deepen the understanding of microscopic properties of solitons and their interaction mechanisms, and moreover provide a promising all-optical solution for the design of eigenvalue-based multiplexers and demultiplexers.

physics.optics↗

arXiv2024

Apollo: An Exploration of Video Understanding in Large Multimodal Models

Despite the rapid integration of video perception capabilities into Large Multimodal Models (LMMs), the underlying mechanisms driving their video understanding remain poorly understood. Consequently, many design decisions in this domain are made without proper justification or analysis. The high computational cost of training and evaluating such models, coupled with limited open research, hinders the development of video-LMMs. To address this, we present a comprehensive study that helps uncover what effectively drives video understanding in LMMs. We begin by critically examining the primary contributors to the high computational requirements associated with video-LMM research and discover Scaling Consistency, wherein design and training decisions made on smaller models and datasets (up to a critical size) effectively transfer to larger models. Leveraging these insights, we explored many video-specific aspects of video-LMMs, including video sampling, architectures, data composition, training schedules, and more. For example, we demonstrated that fps sampling during training is vastly preferable to uniform frame sampling and which vision encoders are the best for video representation. Guided by these findings, we introduce Apollo, a state-of-the-art family of LMMs that achieve superior performance across different model sizes. Our models can perceive hour-long videos efficiently, with Apollo-3B outperforming most existing $7$B models with an impressive 55.1 on LongVideoBench. Apollo-7B is state-of-the-art compared to 7B LMMs with a 70.9 on MLVU, and 63.3 on Video-MME.

cs.CV↗

arXiv2024

Accelerating Multimodal Large Language Models by Searching Optimal Vision Token Reduction

Prevailing Multimodal Large Language Models (MLLMs) encode the input image(s) as vision tokens and feed them into the language backbone, similar to how Large Language Models (LLMs) process the text tokens. However, the number of vision tokens increases quadratically as the image resolutions, leading to huge computational costs. In this paper, we consider improving MLLM's efficiency from two scenarios, (I) Reducing computational cost without degrading the performance. (II) Improving the performance with given budgets. We start with our main finding that the ranking of each vision token sorted by attention scores is similar in each layer except the first layer. Based on it, we assume that the number of essential top vision tokens does not increase along layers. Accordingly, for Scenario I, we propose a greedy search algorithm (G-Search) to find the least number of vision tokens to keep at each layer from the shallow to the deep. Interestingly, G-Search is able to reach the optimal reduction strategy based on our assumption. For Scenario II, based on the reduction strategy from G-Search, we design a parametric sigmoid function (P-Sigmoid) to guide the reduction at each layer of the MLLM, whose parameters are optimized by Bayesian Optimization. Extensive experiments demonstrate that our approach can significantly accelerate those popular MLLMs, e.g. LLaVA, and InternVL2 models, by more than $2 \times$ without performance drops. Our approach also far outperforms other token reduction methods when budgets are limited, achieving a better trade-off between efficiency and effectiveness.

cs.CV↗

arXiv2023

ControlRoom3D: Room Generation using Semantic Proxy Rooms

Manually creating 3D environments for AR/VR applications is a complex process requiring expert knowledge in 3D modeling software. Pioneering works facilitate this process by generating room meshes conditioned on textual style descriptions. Yet, many of these automatically generated 3D meshes do not adhere to typical room layouts, compromising their plausibility, e.g., by placing several beds in one bedroom. To address these challenges, we present ControlRoom3D, a novel method to generate high-quality room meshes. Central to our approach is a user-defined 3D semantic proxy room that outlines a rough room layout based on semantic bounding boxes and a textual description of the overall room style. Our key insight is that when rendered to 2D, this 3D representation provides valuable geometric and semantic information to control powerful 2D models to generate 3D consistent textures and geometry that aligns well with the proxy room. Backed up by an extensive study including quantitative metrics and qualitative user evaluations, our method generates diverse and globally plausible 3D room meshes, thus empowering users to design 3D rooms effortlessly without specialized knowledge.

cs.CV↗

arXiv2023

Emu: Enhancing Image Generation Models Using Photogenic Needles in a Haystack

Training text-to-image models with web scale image-text pairs enables the generation of a wide range of visual concepts from text. However, these pre-trained models often face challenges when it comes to generating highly aesthetic images. This creates the need for aesthetic alignment post pre-training. In this paper, we propose quality-tuning to effectively guide a pre-trained model to exclusively generate highly visually appealing images, while maintaining generality across visual concepts. Our key insight is that supervised fine-tuning with a set of surprisingly small but extremely visually appealing images can significantly improve the generation quality. We pre-train a latent diffusion model on $1.1$ billion image-text pairs and fine-tune it with only a few thousand carefully selected high-quality images. The resulting model, Emu, achieves a win rate of $82.9\%$ compared with its pre-trained only counterpart. Compared to the state-of-the-art SDXLv1.0, Emu is preferred $68.4\%$ and $71.3\%$ of the time on visual appeal on the standard PartiPrompts and our Open User Input benchmark based on the real-world usage of text-to-image models. In addition, we show that quality-tuning is a generic approach that is also effective for other architectures, including pixel diffusion and masked generative transformer models.

cs.CV↗

arXiv2023

Cost-Aware Evaluation and Model Scaling for LiDAR-Based 3D Object Detection

Considerable research effort has been devoted to LiDAR-based 3D object detection and empirical performance has been significantly improved. While progress has been encouraging, we observe an overlooked issue: it is not yet common practice to compare different 3D detectors under the same cost, e.g., inference latency. This makes it difficult to quantify the true performance gain brought by recently proposed architecture designs. The goal of this work is to conduct a cost-aware evaluation of LiDAR-based 3D object detectors. Specifically, we focus on SECOND, a simple grid-based one-stage detector, and analyze its performance under different costs by scaling its original architecture. Then we compare the family of scaled SECOND with recent 3D detection methods, such as Voxel R-CNN and PV-RCNN++. The results are surprising. We find that, if allowed to use the same latency, SECOND can match the performance of PV-RCNN++, the current state-of-the-art method on the Waymo Open Dataset. Scaled SECOND also easily outperforms many recent 3D detection methods published during the past year. We recommend future research control the inference cost in their empirical comparison and include the family of scaled SECOND as a strong baseline when presenting novel 3D detection methods.

cs.CV↗

arXiv2022

Learning Multi-Task Gaussian Process Over Heterogeneous Input Domains

Multi-task Gaussian process (MTGP) is a well-known non-parametric Bayesian model for learning correlated tasks effectively by transferring knowledge across tasks. But current MTGPs are usually limited to the multi-task scenario defined in the same input domain, leaving no space for tackling the heterogeneous case, i.e., the features of input domains vary over tasks. To this end, this paper presents a novel heterogeneous stochastic variational linear model of coregionalization (HSVLMC) model for simultaneously learning the tasks with varied input domains. Particularly, we develop the stochastic variational framework with Bayesian calibration that (i) takes into account the effect of dimensionality reduction raised by domain mappings in order to achieve effective input alignment; and (ii) employs a residual modeling strategy to leverage the inductive bias brought by prior domain mappings for better model inference. Finally, the superiority of the proposed model against existing LMC models has been extensively verified on diverse heterogeneous multi-task cases and a practical multi-fidelity steam turbine exhaust problem.

stat.ML↗

arXiv2022

Deep Probabilistic Time Series Forecasting using Augmented Recurrent Input for Dynamic Systems

The demand of probabilistic time series forecasting has been recently raised in various dynamic system scenarios, for example, system identification and prognostic and health management of machines. To this end, we combine the advances in both deep generative models and state space model (SSM) to come up with a novel, data-driven deep probabilistic sequence model. Specifically, we follow the popular encoder-decoder generative structure to build the recurrent neural networks (RNN) assisted variational sequence model on an augmented recurrent input space, which could induce rich stochastic sequence dependency. Besides, in order to alleviate the inconsistency issue of the posterior between training and predicting as well as improving the mining of dynamic patterns, we (i) propose using a lagged hybrid output as input for the posterior at next time step, which brings training and predicting into alignment; and (ii) further devise a generalized auto-regressive strategy that encodes all the historical dependencies for the posterior. Thereafter, we first investigate the methodological characteristics of the proposed deep probabilistic sequence model on toy cases, and then comprehensively demonstrate the superiority of our model against existing deep probabilistic SSM models through extensive numerical experiments on eight system identification benchmarks from various dynamic systems. Finally, we apply our sequence model to a real-world centrifugal compressor forecasting problem, and again verify its outstanding performance by quantifying the time series predictive distribution.

cs.LG↗

arXiv2022

Wisdom of Committees: An Overlooked Approach To Faster and More Accurate Models

Committee-based models (ensembles or cascades) construct models by combining existing pre-trained ones. While ensembles and cascades are well-known techniques that were proposed before deep learning, they are not considered a core building block of deep model architectures and are rarely compared to in recent literature on developing efficient models. In this work, we go back to basics and conduct a comprehensive analysis of the efficiency of committee-based models. We find that even the most simplistic method for building committees from existing, independently pre-trained models can match or exceed the accuracy of state-of-the-art models while being drastically more efficient. These simple committee-based models also outperform sophisticated neural architecture search methods (e.g., BigNAS). These findings hold true for several tasks, including image classification, video classification, and semantic segmentation, and various architecture families, such as ViT, EfficientNet, ResNet, MobileNetV2, and X3D. Our results show that an EfficientNet cascade can achieve a 5.4x speedup over B7 and a ViT cascade can achieve a 2.3x speedup over ViT-L-384 while being equally accurate.

cs.CV↗

arXiv2021

Neighborhood-Aware Neural Architecture Search

Existing neural architecture search (NAS) methods often return an architecture with good search performance but generalizes poorly to the test setting. To achieve better generalization, we propose a novel neighborhood-aware NAS formulation to identify flat-minima architectures in the search space, with the assumption that flat minima generalize better than sharp minima. The phrase ``flat-minima architecture'' refers to architectures whose performance is stable under small perturbations in the architecture (e.g., replacing a convolution with a skip connection). Our formulation takes the ``flatness'' of an architecture into account by aggregating the performance over the neighborhood of this architecture. We demonstrate a principled way to apply our formulation to existing search algorithms, including sampling-based algorithms and gradient-based algorithms. To facilitate the application to gradient-based algorithms, we also propose a differentiable representation for the neighborhood of architectures. Based on our formulation, we propose neighborhood-aware random search (NA-RS) and neighborhood-aware differentiable architecture search (NA-DARTS). Notably, by simply augmenting DARTS with our formulation, NA-DARTS outperforms DARTS and achieves state-of-the-art performance on established benchmarks, including CIFAR-10, CIFAR-100 and ImageNet.

cs.LG↗

arXiv2021

Scalable Multi-Task Gaussian Processes with Neural Embedding of Coregionalization

Multi-task regression attempts to exploit the task similarity in order to achieve knowledge transfer across related tasks for performance improvement. The application of Gaussian process (GP) in this scenario yields the non-parametric yet informative Bayesian multi-task regression paradigm. Multi-task GP (MTGP) provides not only the prediction mean but also the associated prediction variance to quantify uncertainty, thus gaining popularity in various scenarios. The linear model of coregionalization (LMC) is a well-known MTGP paradigm which exploits the dependency of tasks through linear combination of several independent and diverse GPs. The LMC however suffers from high model complexity and limited model capability when handling complicated multi-task cases. To this end, we develop the neural embedding of coregionalization that transforms the latent GPs into a high-dimensional latent space to induce rich yet diverse behaviors. Furthermore, we use advanced variational inference as well as sparse approximation to devise a tight and compact evidence lower bound (ELBO) for higher quality of scalable model inference. Extensive numerical experiments have been conducted to verify the higher prediction quality and better generalization of our model, named NSVLMC, on various real-world multi-task datasets and the cross-fluid modeling of unsteady fluidized bed.

stat.ML↗

arXiv2021

Efficient Model Performance Estimation via Feature Histories

An important step in the task of neural network design, such as hyper-parameter optimization (HPO) or neural architecture search (NAS), is the evaluation of a candidate model's performance. Given fixed computational resources, one can either invest more time training each model to obtain more accurate estimates of final performance, or spend more time exploring a greater variety of models in the configuration space. In this work, we aim to optimize this exploration-exploitation trade-off in the context of HPO and NAS for image classification by accurately approximating a model's maximal performance early in the training process. In contrast to recent accelerated NAS methods customized for certain search spaces, e.g., requiring the search space to be differentiable, our method is flexible and imposes almost no constraints on the search space. Our method uses the evolution history of features of a network during the early stages of training to build a proxy classifier that matches the peak performance of the network under consideration. We show that our method can be combined with multiple search algorithms to find better solutions to a wide range of tasks in HPO and NAS. Using a sampling-based search algorithm and parallel computing, our method can find an architecture which is better than DARTS and with an 80% reduction in wall-clock search time.

cs.LG↗

arXiv2020

Modulating Scalable Gaussian Processes for Expressive Statistical Learning

For a learning task, Gaussian process (GP) is interested in learning the statistical relationship between inputs and outputs, since it offers not only the prediction mean but also the associated variability. The vanilla GP however struggles to learn complicated distribution with the property of, e.g., heteroscedastic noise, multi-modality and non-stationarity, from massive data due to the Gaussian marginal and the cubic complexity. To this end, this article studies new scalable GP paradigms including the non-stationary heteroscedastic GP, the mixture of GPs and the latent GP, which introduce additional latent variables to modulate the outputs or inputs in order to learn richer, non-Gaussian statistical representation. We further resort to different variational inference strategies to arrive at analytical or tighter evidence lower bounds (ELBOs) of the marginal likelihood for efficient and effective model training. Extensive numerical experiments against state-of-the-art GP and neural network (NN) counterparts on various tasks verify the superiority of these scalable modulated GPs, especially the scalable latent GP, for learning diverse data distributions.

stat.ML↗

arXiv2020

Deep Latent-Variable Kernel Learning

Deep kernel learning (DKL) leverages the connection between Gaussian process (GP) and neural networks (NN) to build an end-to-end, hybrid model. It combines the capability of NN to learn rich representations under massive data and the non-parametric property of GP to achieve automatic regularization that incorporates a trade-off between model fit and model complexity. However, the deterministic encoder may weaken the model regularization of the following GP part, especially on small datasets, due to the free latent representation. We therefore present a complete deep latent-variable kernel learning (DLVKL) model wherein the latent variables perform stochastic encoding for regularized representation. We further enhance the DLVKL from two aspects: (i) the expressive variational posterior through neural stochastic differential equation (NSDE) to improve the approximation quality, and (ii) the hybrid prior taking knowledge from both the SDE prior and the posterior to arrive at a flexible trade-off. Intensive experiments imply that the DLVKL-NSDE performs similarly to the well calibrated GP on small datasets, and outperforms existing deep GPs on large datasets.

stat.ML↗

arXiv2020

AttentionNAS: Spatiotemporal Attention Cell Search for Video Classification

Convolutional operations have two limitations: (1) do not explicitly model where to focus as the same filter is applied to all the positions, and (2) are unsuitable for modeling long-range dependencies as they only operate on a small neighborhood. While both limitations can be alleviated by attention operations, many design choices remain to be determined to use attention, especially when applying attention to videos. Towards a principled way of applying attention to videos, we address the task of spatiotemporal attention cell search. We propose a novel search space for spatiotemporal attention cells, which allows the search algorithm to flexibly explore various design choices in the cell. The discovered attention cells can be seamlessly inserted into existing backbone networks, e.g., I3D or S3D, and improve video classification accuracy by more than 2% on both Kinetics-600 and MiT datasets. The discovered attention cells outperform non-local blocks on both datasets, and demonstrate strong generalization across different modalities, backbones, and datasets. Inserting our attention cells into I3D-R50 yields state-of-the-art performance on both datasets.

cs.CV↗

arXiv2019

Learnable Embedding Space for Efficient Neural Architecture Compression

We propose a method to incrementally learn an embedding space over the domain of network architectures, to enable the careful selection of architectures for evaluation during compressed architecture search. Given a teacher network, we search for a compressed network architecture by using Bayesian Optimization (BO) with a kernel function defined over our proposed embedding space to select architectures for evaluation. We demonstrate that our search algorithm can significantly outperform various baseline methods, such as random search and reinforcement learning (Ashok et al., 2018). The compressed architectures found by our method are also better than the state-of-the-art manually-designed compact architecture ShuffleNet (Zhang et al., 2018). We also demonstrate that the learned embedding space can be transferred to new settings for architecture search, such as a larger teacher network or a teacher network in a different architecture family, without any training. Code is publicly available here: https://github.com/Friedrich1006/ESNAC .

cs.CV↗