arXiv subjects

Xiaoxu Chen

Publications and source records attributed to Xiaoxu Chen.

15 recordsLinked to original sources

arXiv2026

HYPIC: Accelerating Hybrid-Attention LLM Serving with Position-Independent Caching

In retrieval-augmented generation and agentic LLM serving, prompts are assembled from independent segments into long contexts, making the prefill stage dominate per-request cost. Two directions have emerged to reduce this cost: position-independent caching (PIC) admits KV reuse for non-contiguous segments shared across requests, while hybrid-attention models cut computation by replacing most full-attention layers with linear attention. However, they cannot coexist: applying existing PIC methods to hybrid-attention models breaks down because per-token KV-cache reuse primitives do not transfer to the per-request recurrent state. We present Hypic, the first system to accelerate hybrid-attention LLM serving with position-independent caching. For linear-attention layers, we identify the segment-cumulative transition operator as the missing algebraic primitive and cache it alongside each segment's zero-start end-state, enabling near-exact and constant-time composition of independently cached segments. For the remaining full-attention layers, existing PIC methods also fail because linear layers do not expose the per-token hidden states needed for selective recomputation. We show that the largest deviations concentrate at segment beginnings and construct a small seam window that propagates hidden states through the hybrid-attention stack to repair cross-segment attention. Finally, Hypic introduces segment parallelism, which exploits PIC's segment-level self-containment to parallelize cache-miss prefill across instances, turning long cold requests into an accelerable workload. Evaluated across four hybrid-attention models and five workloads, Hypic reduces time-to-first-token by $3.25\times$ on average and improves QPS by $1.66\times$ over Prefix Cache, while preserving task quality with a 1.71-point gap from Full Recompute.

cs.DC↗

arXiv2026

Vivid-VR: Distilling Concepts from Text-to-Video Diffusion Transformer for Photorealistic Video Restoration

We present Vivid-VR, a DiT-based generative video restoration method built upon an advanced T2V foundation model, where ControlNet is leveraged to control the generation process, ensuring content consistency. However, conventional fine-tuning of such controllable pipelines frequently suffers from distribution drift due to limitations in imperfect multimodal alignment, resulting in compromised texture realism and temporal coherence. To tackle this challenge, we propose a concept distillation training strategy that utilizes the pretrained T2V model to synthesize training samples with embedded textual concepts, thereby distilling its conceptual understanding to preserve texture and temporal quality. To enhance generation controllability, we redesign the control architecture with two key components: 1) a control feature projector that filters degradation artifacts from input video latents to minimize their propagation through the generation pipeline, and 2) a new ControlNet connector employing a dual-branch design. This connector synergistically combines MLP-based feature mapping with cross-attention mechanism for dynamic control feature retrieval, enabling both content preservation and adaptive control signal modulation. Extensive experiments show that Vivid-VR performs favorably against existing approaches on both synthetic and real-world benchmarks, as well as AIGC videos, achieving impressive texture realism, visual vividness, and temporal consistency. The codes and checkpoints are publicly available at https://github.com/csbhr/Vivid-VR.

cs.CV↗

arXiv2026

SATB-VR: Training Few-Step Video Restoration Diffusion Model using SNR-Aware Trajectory Blending

While diffusion models excel in video restoration, their reliance on extensive iterative steps limits efficiency. Conversely, aggressive single-step distillation often compromises fine texture recovery. To achieve an optimal balance, we present SATB-VR, a few-step paradigm that jump-starts the denoising process via an auxiliary predictor, explicitly bypassing early low signal-to-noise ratio (SNR) steps. However, naive joint training of the predictor and the denoiser inherently introduces a severe train-inference discrepancy. To resolve this, we propose the SNR-Aware Trajectory Blending (SATB) strategy. During the forward process, SATB constructs the noisy input by dynamically blending the predictor's output with the ground-truth trajectory based on the SNRs. This forces the denoiser to robustly compensate for initial prediction errors while smoothly converging to the clean data manifold. Furthermore, we introduce a Denoiser-Driven Consistency (DDC) loss, leveraging the concurrently updated denoiser as a dynamic evaluator to explicitly align internal features and boost predictor accuracy. Extensive experiments demonstrate that, under flexible few-step inference regimes (\eg, $\le 5$ steps), SATB-VR performs favorably against existing approaches on synthetic, real-world, and AIGC benchmarks.

cs.CV↗

arXiv2026

NTIRE 2026 Challenge on Short-form UGC Video Restoration in the Wild with Generative Models: Datasets, Methods and Results

This paper presents an overview of the NTIRE 2026 Challenge on Short-form UGC Video Restoration in the Wild with Generative Models. This challenge utilizes a new short-form UGC (S-UGC) video restoration benchmark, termed KwaiVIR, which is contributed by USTC and Kuaishou Technology. It contains both synthetically distorted videos and real-world short-form UGC videos in the wild. For this edition, the released data include 200 synthetic training videos, 48 wild training videos, 11 validation videos, and 20 testing videos. The primary goal of this challenge is to establish a strong and practical benchmark for restoring short-form UGC videos under complex real-world degradations, especially in the emerging paradigm of generative-model-based S-UGC video restoration. This challenge has two tracks: (i) the primary track is a subjective track, where the evaluation is based on a user study; (ii) the second track is an objective track. These two tracks enable a comprehensive assessment of restoration quality. In total, 95 teams have registered for this competition. And 12 teams submitted valid final solutions and fact sheets for the testing phase. The submitted methods achieved strong performance on the KwaiVIR benchmark, demonstrating encouraging progress in short-form UGC video restoration in the wild.

cs.CV↗

arXiv2025

Bayesian spatiotemporal modeling of passenger trip assignment in metro networks

Assigning passenger trips to specific network paths using automatic fare collection (AFC) data is a fundamental application in urban transit analysis. The task is a difficult inverse problem: the only available information consists of each passenger's total travel time and their origin and destination, while individual passenger path choices and dynamic network costs are unobservable, and behavior varies significantly across space and time. We propose a novel Bayesian hierarchical model to resolve this problem by jointly estimating dynamic network costs and passenger path choices while quantifying their uncertainty. Our model decomposes trip travel time into four components -- access, in-vehicle, transfer, and egress -- each modeled as a time-varying random walk. To capture heterogeneous passenger behavior, we introduce a multinomial logit model with spatiotemporally varying coefficients. We manage the high dimensionality of these coefficients using kernelized tensor factorization with Gaussian process priors to effectively model complex spatiotemporal correlations. We develop a tailored and efficient Markov chain Monte Carlo (MCMC) algorithm for model inference. A simulation study demonstrates the method's effectiveness in recovering the underlying model parameters. On a large-scale dataset from the Hong Kong Mass Transit Railway, our framework demonstrates superior estimation accuracy over established benchmarks. The results reveal significant spatiotemporal variations in passenger preferences and provide robust uncertainty quantification, offering transit operators a powerful tool for enhancing service planning and operational management.

stat.AP↗

arXiv2025

YingLong-weather: AI-Based Limited Area Models for Forecasting of Non-precipitation Surface Meteorological Variables

Recently, artificial intelligence-based (AI-based) models for forecasting of global weather have been rapidly developed. Most of the global models are trained on reanalysis datasets with a spatial resolution of 0.25°*0.25°. However, research on AI-based high spatial resolution limited area weather forecasting models remains limited. In this study, YingLong, an AI-based limited area weather forecasting model with a spatial resolution of 3 km * 3 km is developed. YingLong employs a parallel structure of global and local blocks to capture multiscale meteorological features and operates much faster than the dynamical limited area model WRF-ARW. In two selected limited areas (one relatively flat and the other featuring significant mountain ranges), YingLong (with lateral boundary condition imposed by the global AI-based model Pangu-weather) demonstrates superior skill in forecasting surface wind speed compared to WRF-ARW. Additionally, it shows comparable skill in forecasting surface temperature and pressure. The accuracy of surface temperature and humidity forecasts can be further improved by applying better boundary conditions. YingLong also addresses issues related to the lateral boundary conditions of AI-based limited area models, such as selecting the width of the lateral boundary region and combining finer and coarser resolution predictions in this region. Therefore, YingLong has a great potential to generate cost-effective multiyear high-resolution synthetic wind speed that maintain meteorological realism both spatially and temporally, aiding in the planning and operations for wind power generation companies.

physics.ao-ph↗

arXiv2024

Bayesian Inference of Time-Varying Origin-Destination Matrices from Boarding/Alighting Counts for Transit Services

Origin-destination (OD) demand matrices are crucial for transit agencies to design and operate transit systems. This paper presents a novel temporal Bayesian model designed to estimate transit OD matrices at the individual bus-journey level from boarding/alighting counts at bus stops. Our approach begins by modeling the number of alighting passengers at subsequent bus stops, given a boarding stop, through a multinomial distribution parameterized by alighting probabilities. Given the large scale of the problem, we generate alighting probabilities with a latent variable matrix and factorize it into a mapping matrix and a temporal matrix, thereby substantially reducing the number of parameters. To further encode a temporally-smooth structure in the parameters, we impose a Gaussian process prior on the columns of the temporal factor matrix. For model inference, we develop a two-stage algorithm with the Markov chain Monte Carlo (MCMC) method. In the first stage, latent OD matrices are sampled conditional on model parameters using a Metropolis-Hastings sampling algorithm with a Markov model-based proposal distribution. In the second stage, we sample model parameters conditional on latent OD matrices using slice and elliptical slice sampling algorithms. We assess the proposed model using real-world data collected from three bus routes with varying numbers of stops, and the results demonstrate that our model achieves accurate posterior mean estimation and outperforms the widely used iterative proportional fitting (IPF) method. Additionally, our model can provide uncertainty quantification for the OD demand matrices, thus benefiting many downstream planning/operational tasks that require robust decisions.

stat.AP↗

arXiv2024

ForecastGrapher: Redefining Multivariate Time Series Forecasting with Graph Neural Networks

The challenge of effectively learning inter-series correlations for multivariate time series forecasting remains a substantial and unresolved problem. Traditional deep learning models, which are largely dependent on the Transformer paradigm for modeling long sequences, often fail to integrate information from multiple time series into a coherent and universally applicable model. To bridge this gap, our paper presents ForecastGrapher, a framework reconceptualizes multivariate time series forecasting as a node regression task, providing a unique avenue for capturing the intricate temporal dynamics and inter-series correlations. Our approach is underpinned by three pivotal steps: firstly, generating custom node embeddings to reflect the temporal variations within each series; secondly, constructing an adaptive adjacency matrix to encode the inter-series correlations; and thirdly, augmenting the GNNs' expressive power by diversifying the node feature distribution. To enhance this expressive power, we introduce the Group Feature Convolution GNN (GFC-GNN). This model employs a learnable scaler to segment node features into multiple groups and applies one-dimensional convolutions with different kernel lengths to each group prior to the aggregation phase. Consequently, the GFC-GNN method enriches the diversity of node feature distribution in a fully end-to-end fashion. Through extensive experiments and ablation studies, we show that ForecastGrapher surpasses strong baselines and leading published techniques in the domain of multivariate time series forecasting.

cs.LG↗

arXiv2024

Towards Real-World Blind Face Restoration with Generative Diffusion Prior

Blind face restoration is an important task in computer vision and has gained significant attention due to its wide-range applications. Previous works mainly exploit facial priors to restore face images and have demonstrated high-quality results. However, generating faithful facial details remains a challenging problem due to the limited prior knowledge obtained from finite data. In this work, we delve into the potential of leveraging the pretrained Stable Diffusion for blind face restoration. We propose BFRffusion which is thoughtfully designed to effectively extract features from low-quality face images and could restore realistic and faithful facial details with the generative prior of the pretrained Stable Diffusion. In addition, we build a privacy-preserving face dataset called PFHQ with balanced attributes like race, gender, and age. This dataset can serve as a viable alternative for training blind face restoration networks, effectively addressing privacy and bias concerns usually associated with the real face datasets. Through an extensive series of experiments, we demonstrate that our BFRffusion achieves state-of-the-art performance on both synthetic and real-world public testing datasets for blind face restoration and our PFHQ dataset is an available resource for training blind face restoration networks. The codes, pretrained models, and dataset are released at https://github.com/chenxx89/BFRffusion.

cs.CV↗

arXiv2024

Conditional forecasting of bus travel time and passenger occupancy with Bayesian Markov regime-switching vector autoregression

Accurately forecasting bus travel time and passenger occupancy with uncertainty is essential for both travelers and transit agencies/operators. However, existing approaches to forecasting bus travel time and passenger occupancy mainly rely on deterministic models, providing only point estimates. In this paper, we develop a Bayesian Markov regime-switching vector autoregressive model to jointly forecast both bus travel time and passenger occupancy with uncertainty. The proposed approach naturally captures the intricate interactions among adjacent buses and adapts to the multimodality and skewness of real-world bus travel time and passenger occupancy observations. We develop an efficient Markov chain Monte Carlo (MCMC) sampling algorithm to approximate the resultant joint posterior distribution of the parameter vector. With this framework, the estimation of downstream bus travel time and passenger occupancy is transformed into a multivariate time series forecasting problem conditional on partially observed outcomes. Experimental validation using real-world data demonstrates the superiority of our proposed model in terms of both predictive means and uncertainty quantification compared to the Bayesian Gaussian mixture model.

stat.AP↗

arXiv2023

Blind Face Restoration for Under-Display Camera via Dictionary Guided Transformer

By hiding the front-facing camera below the display panel, Under-Display Camera (UDC) provides users with a full-screen experience. However, due to the characteristics of the display, images taken by UDC suffer from significant quality degradation. Methods have been proposed to tackle UDC image restoration and advances have been achieved. There are still no specialized methods and datasets for restoring UDC face images, which may be the most common problem in the UDC scene. To this end, considering color filtering, brightness attenuation, and diffraction in the imaging process of UDC, we propose a two-stage network UDC Degradation Model Network named UDC-DMNet to synthesize UDC images by modeling the processes of UDC imaging. Then we use UDC-DMNet and high-quality face images from FFHQ and CelebA-Test to create UDC face training datasets FFHQ-P/T and testing datasets CelebA-Test-P/T for UDC face restoration. We propose a novel dictionary-guided transformer network named DGFormer. Introducing the facial component dictionary and the characteristics of the UDC image in the restoration makes DGFormer capable of addressing blind face restoration in UDC scenarios. Experiments show that our DGFormer and UDC-DMNet achieve state-of-the-art performance.

cs.CV↗

arXiv2023

Enhancing Representation Learning for Periodic Time Series with Floss: A Frequency Domain Regularization Approach

Time series analysis is a fundamental task in various application domains, and deep learning approaches have demonstrated remarkable performance in this area. However, many real-world time series data exhibit significant periodic or quasi-periodic dynamics that are often not adequately captured by existing deep learning-based solutions. This results in an incomplete representation of the underlying dynamic behaviors of interest. To address this gap, we propose an unsupervised method called Floss that automatically regularizes learned representations in the frequency domain. The Floss method first automatically detects major periodicities from the time series. It then employs periodic shift and spectral density similarity measures to learn meaningful representations with periodic consistency. In addition, Floss can be easily incorporated into both supervised, semi-supervised, and unsupervised learning frameworks. We conduct extensive experiments on common time series classification, forecasting, and anomaly detection tasks to demonstrate the effectiveness of Floss. We incorporate Floss into several representative deep learning solutions to justify our design choices and demonstrate that it is capable of automatically discovering periodic dynamics and improving state-of-the-art deep learning models.

cs.LG↗

arXiv2022

Discovering Dynamic Patterns from Spatiotemporal Data with Time-Varying Low-Rank Autoregression

The problem of broad practical interest in spatiotemporal data analysis, i.e., discovering interpretable dynamic patterns from spatiotemporal data, is studied in this paper. Towards this end, we develop a time-varying reduced-rank vector autoregression (VAR) model whose coefficient matrices are parameterized by low-rank tensor factorization. Benefiting from the tensor factorization structure, the proposed model can simultaneously achieve model compression and pattern discovery. In particular, the proposed model allows one to characterize nonstationarity and time-varying system behaviors underlying spatiotemporal data. To evaluate the proposed model, extensive experiments are conducted on various spatiotemporal data representing different nonlinear dynamical systems, including fluid dynamics, sea surface temperature, USA surface temperature, and NYC taxi trips. Experimental results demonstrate the effectiveness of modeling spatiotemporal data and characterizing spatial/temporal patterns with the proposed model. In the spatial context, the spatial patterns can be automatically extracted and intuitively characterized by the spatial modes. In the temporal context, the complex time-varying system behaviors can be revealed by the temporal modes in the proposed model. Thus, our model lays an insightful foundation for understanding complex spatiotemporal data in real-world dynamical systems. The adapted datasets and Python implementation are publicly available at https://github.com/xinychen/vars.

cs.LG↗

arXiv2022

Probabilistic forecasting of bus travel time with a Bayesian Gaussian mixture model

Accurate forecasting of bus travel time and its uncertainty is critical to service quality and operation of transit systems; for example, it can help passengers make better decisions on departure time, route choice, and even transport mode choice and also support transit operators to make informed decisions on tasks such as crew/vehicle scheduling and timetabling. However, most existing approaches in bus travel time forecasting are based on deterministic models that provide only point estimation. To this end, we develop in this paper a Bayesian probabilistic forecasting model for bus travel time. To characterize the strong dependencies/interactions between consecutive buses, we concatenate the link travel time vectors and the headway vector from a pair of two adjacent buses as a new augmented variable and model it with a constrained Multivariate Gaussian mixture distributions. This approach can naturally capture the interactions between adjacent buses (e.g., correlated speed and smooth variation of headway), handle missing values in data, and depict the multimodality in bus travel time distributions. Next, we assume different periods in a day share the same set of Gaussian components but different mixing coefficients to characterize the systematic temporal variations in bus operation. For model inference, we develop an efficient Markov chain Monte Carlo (MCMC) sampling algorithm to obtain the posterior distributions of model parameters and make probabilistic forecasting. We test the proposed model using the data from a twenty-link bus route in Guangzhou, China. Results show our approach significantly outperforms baseline models that overlook bus-to-bus interactions in terms of both predictive means and distributions. Besides forecasting, the parameters of the proposed model contain rich information for understanding/improving the bus service.

stat.AP↗

arXiv2022

Bayesian inference for link travel time correlation of a bus route

Estimation of link travel time correlation of a bus route is essential to many bus operation applications, such as timetable scheduling, travel time forecasting and transit service assessment/improvement. Most previous studies rely on either independent assumptions or simplified local spatial correlation structures. In the real world, however, link travel time on a bus route could exhibit complex correlation structures, such as long-range correlations, negative correlations, and time-varying correlations. Therefore, before introducing strong assumptions, it is essential to empirically quantify and examine the correlation structure of link travel time from real-world bus operation data. To this end, this paper develops a Bayesian Gaussian model to estimate the link travel time correlation matrix of a bus route using smart-card-like data. Our method overcomes the small-sample-size problem in correlation matrix estimation by borrowing/integrating those incomplete observations (i.e., with missing/ragged values and overlapped link segments) from other bus routes. Next, we propose an efficient Gibbs sampling framework to marginalize over the missing and ragged values and obtain the posterior distribution of the correlation matrix. Three numerical experiments are conducted to evaluate model performance. We first conduct a synthetic experiment and our results show that the proposed method produces an accurate estimation for travel time correlations with credible intervals. Next, we perform experiments on a real-world bus route with smart card data; our results show that both local and long-range correlations exist on this bus route. Finally, we demonstrate an application of using the estimated covariance matrix to make probabilistic forecasting of link and trip travel time.

stat.AP↗