arXiv subjects

Xi Weng

Publications and source records attributed to Xi Weng.

16 recordsLinked to original sources

arXiv2026

Contextual Robust Optimization for AI Data Center Scheduling with Statistical Guarantees

The rapid growth of AI workloads is substantially increasing data center electricity demand and carbon emissions, motivating the development of carbon-aware scheduling methods. However, effective scheduling is challenging because renewable generation and AI workloads are subject to forecast errors, while training and inference workloads exhibit heterogeneity in computational characteristics. This paper proposes a contextual robust optimization framework for AI data center operation. The proposed model explicitly captures the heterogeneous computational characteristics of AI training and inference workloads. To deal with renewable generation and workload forecast errors, we develop loss-based uncertainty learning models that directly map contextual features to covariate-dependent uncertainty sets. The resulting contextual joint chance-constrained scheduling problem is reformulated into a tractable robust optimization problem, and a calibration algorithm is developed to provide finite-sample probabilistic feasibility guarantees for multiple joint chance constraints. Numerical experiments based on real-world AI workload traces and renewable generation data show that the proposed method reduces operating costs by an average of 5.57% compared to benchmark methods while maintaining reliable feasibility and strong computational scalability.

math.OC↗

arXiv2026

Mechanism and Communication Co-Design for Differentially Private Energy Sharing

Integrating distributed energy resources (DERs) is a critical step toward addressing the global climate crisis. This transformation has driven the transition from traditional consumers to prosumers and given rise to new energy sharing business models. Existing works have extensively studied prosumer energy sharing mechanisms, yet little attention has been paid to privacy protection, particularly when communication constraints are taken into account. In this paper, we study an energy sharing mechanism where information is exchanged over wireless channels via over-the-air (OTA) multiple-input multiple-output (MIMO) aggregation to exploit spectral efficiency for scalable prosumer coordination. To characterize the privacy leakage risk during data transmission process, we introduce an adversarial attack model and demonstrate that, under certain conditions, the platform can extract and recover prosumers' private parameters from the base station observations. To safeguard the energy sharing mechanism against such attacks, we propose a differentially private equilibrium-seeking algorithm, analyze the achievable privacy level, and establish convergence guarantees that quantify the impact of privacy on the convergence accuracy. Numerical experiments demonstrate that our approach effectively protects prosumers' privacy while converging to near-optimal solutions.

cs.IT↗

arXiv2026

TennisExpert: Towards Expert-Level Analytical Sports Video Understanding

Tennis is one of the most widely followed sports, generating extensive broadcast footage with strong potential for professional analysis, automated coaching, and real-time commentary. However, automatic tennis understanding remains underexplored due to two key challenges: (1) the lack of large-scale benchmarks with fine-grained annotations and expert-level commentary, and (2) the difficulty of building accurate yet efficient multimodal systems suitable for real-time deployment. To address these challenges, we introduce TennisVL, a large-scale tennis benchmark comprising over 200 professional matches (471.9 hours) and 40,000+ rally-level clips. Unlike existing commentary datasets that focus on descriptive play-by-play narration, TennisVL emphasizes expert analytical commentary capturing tactical reasoning, player decisions, and match momentum. Furthermore, we propose TennisExpert, a multimodal tennis understanding framework that integrates a video semantic parser with a memory-augmented model built on Qwen3-VL-8B. The parser extracts key match elements (e.g., scores, shot sequences, ball bounces, and player locations), while hierarchical memory modules capture both short- and long-term temporal context. Experiments show that TennisExpert consistently outperforms strong proprietary baselines, including GPT-5, Gemini, and Claude, and demonstrates improved ability to capture tactical context and match dynamics. Our dataset and code are publicly available at https://github.com/LZYAndy/TennisExpert.

cs.CV↗

arXiv2025

Clustering Properties of Self-Supervised Learning

Self-supervised learning (SSL) methods via joint embedding architectures have proven remarkably effective at capturing semantically rich representations with strong clustering properties, magically in the absence of label supervision. Despite this, few of them have explored leveraging these untapped properties to improve themselves. In this paper, we provide an evidence through various metrics that the encoder's output $encoding$ exhibits superior and more stable clustering properties compared to other components. Building on this insight, we propose a novel positive-feedback SSL method, termed Representation Self-Assignment (ReSA), which leverages the model's clustering properties to promote learning in a self-guided manner. Extensive experiments on standard SSL benchmarks reveal that models pretrained with ReSA outperform other state-of-the-art SSL methods by a significant margin. Finally, we analyze how ReSA facilitates better clustering properties, demonstrating that it effectively enhances clustering performance at both fine-grained and coarse-grained levels, shaping representations that are inherently more structured and semantically meaningful.

cs.LG↗

arXiv2025

TinyLLaVA-Video: Towards Smaller LMMs for Video Understanding with Group Resampler

Video behavior recognition and scene understanding are fundamental tasks in multimodal intelligence, serving as critical building blocks for numerous real-world applications. Through large multimodal models (LMMs) have achieved remarkable progress in video understanding, most existing open-source models rely on over 7B parameters and require large-scale datasets for training, making them resource-intensive and inaccessible to many researchers. Furthermore, lightweight models face persistent challenges in effectively processing long visual sequences and temporal understanding. In this work, we introduce TinyLLaVA-Video, a lightweight yet powerful video understanding model with approximately 3.6B parameters. The cornerstone of our design is the video-level group resampler, a novel mechanism that significantly reduces and controls the number of visual tokens at the video level. Unlike traditional image-level resampler, our approach effectively mitigates redundancy while enhancing temporal comprehension, leading to improved performance on video-based tasks. In addition, TinyLLaVA-Video demonstrates exceptional efficiency, requiring only one day of training on 8 A100-40G GPUs. It surpasses several existing 7B-parameter models on multiple benchmarks. We believe this work provides a valuable foundation for future research on lightweight video understanding models. The code and weights is available at https://github.com/ZhangXJ199/TinyLLaVA-Video.

cs.CV↗

arXiv2024

Online Low-Carbon Workload, Energy, and Temperature Management of Distributed Data Centers

Data centers have become one of the major energy consumers, making their low-carbon operations critical to achieving global carbon neutrality. Although distributed data centers have the potential to reduce costs and emissions through cooperation, they are facing challenges due to uncertainties. This paper proposes an online approach to co-optimize the workload, energy, and temperature strategies across distributed data centers, targeting minimal total cost, controlled carbon emissions, and adherence to operational constraints. Lyapunov optimization technique is adopted to derive a parametric real-time strategy that accommodates uncertainties in workload demands, ambient temperature, electricity prices, and carbon intensities, without requiring prior knowledge of their distributions. A theoretical upper bound for the optimality gap is derived, based on which a linear programming problem is proposed to optimize the strategy parameters, enhancing performance while ensuring operational constraints. Case studies and method comparison validate the proposed method's effectiveness in reducing costs and carbon emissions.

eess.SY↗

arXiv2024

Meta-Backscatter: A New ISAC Paradigm for Battery-Free Internet of Things

The meta-material sensor has been regarded as a next-generation sensing technology for the battery-free Internet of Things (IoT) due to its battery-free characteristic and improved sensing performance. The meta-material sensors function as backscatter tags that change their reflection coefficients with the conditions of sensing targets such as temperature and gas concentration, allowing transceivers to perform sensing by analyzing the reflected signals from the sensors. Simultaneously, the sensors also function as environmental scatterers, creating additional signal paths to enhance communication performance. Therefore, the meta-material sensor potentially provides a new paradigm of Integrated Sensing and Communication (ISAC) for the battery-free IoT system. In this article, we first propose a Meta-Backscatter system that utilizes meta-material sensors to achieve diverse sensing functionalities and improved communication performance. We begin with the introduction of the metamaterial sensor and further elaborate on the Meta-Backscatter system. Subsequently, we present optimization strategies for meta-material sensors, transmitters, and receivers to strike a balance between sensing and communication. Furthermore, this article provides a case study of the system and examines the feasibility and trade-off through the simulation results. Finally, potential extensions of the system and their related research challenges are addressed.

eess.SY↗

arXiv2024

TinyLLaVA Factory: A Modularized Codebase for Small-scale Large Multimodal Models

We present TinyLLaVA Factory, an open-source modular codebase for small-scale large multimodal models (LMMs) with a focus on simplicity of code implementations, extensibility of new features, and reproducibility of training results. Following the design philosophy of the factory pattern in software engineering, TinyLLaVA Factory modularizes the entire system into interchangeable components, with each component integrating a suite of cutting-edge models and methods, meanwhile leaving room for extensions to more features. In addition to allowing users to customize their own LMMs, TinyLLaVA Factory provides popular training recipes to let users pretrain and finetune their models with less coding effort. Empirical experiments validate the effectiveness of our codebase. The goal of TinyLLaVA Factory is to assist researchers and practitioners in exploring the wide landscape of designing and training small-scale LMMs with affordable computational resources.

cs.LG↗

arXiv2024

TinyLLaVA: A Framework of Small-scale Large Multimodal Models

We present the TinyLLaVA framework that provides a unified perspective in designing and analyzing the small-scale Large Multimodal Models (LMMs). We empirically study the effects of different vision encoders, connection modules, language models, training data and training recipes. Our extensive experiments showed that better quality of data combined with better training recipes, smaller LMMs can consistently achieve on-par performances compared to bigger LMMs. Under our framework, we train a family of small-scale LMMs. Our best model, TinyLLaVA-3.1B, achieves better overall performance against existing 7B models such as LLaVA-1.5 and Qwen-VL. We hope our findings can serve as baselines for future research in terms of data scaling, training setups and model selections. Our model weights and codes will be made public.

cs.LG↗

arXiv2023

Modulate Your Spectrum in Self-Supervised Learning

Whitening loss offers a theoretical guarantee against feature collapse in self-supervised learning (SSL) with joint embedding architectures. Typically, it involves a hard whitening approach, transforming the embedding and applying loss to the whitened output. In this work, we introduce Spectral Transformation (ST), a framework to modulate the spectrum of embedding and to seek for functions beyond whitening that can avoid dimensional collapse. We show that whitening is a special instance of ST by definition, and our empirical investigations unveil other ST instances capable of preventing collapse. Additionally, we propose a novel ST instance named IterNorm with trace loss (INTL). Theoretical analysis confirms INTL's efficacy in preventing collapse and modulating the spectrum of embedding toward equal-eigenvalues during optimization. Our experiments on ImageNet classification and COCO object detection demonstrate INTL's potential in learning superior representations. The code is available at https://github.com/winci-ai/INTL.

cs.LG↗

arXiv2022

Optimal Pricing Schemes in the Presence of Social Learning and Costly Reporting

A monopoly platform sells either a risky product (with unknown utility) or a safe product (with known utility) to agents who sequentially arrive and learn the utility of the risky product by the reporting of previous agents. It is costly for agents to report utility; hence the platform has to design both the prices and the reporting bonus to motivate the agents to explore and generate new information. By allowing sellers to set bonuses, we are essentially enabling them to dynamically control the supply of learning signals without significantly affecting the demand for the product. We characterize the optimal bonus and pricing schemes offered by the profit-maximizing platform. It turns out that the optimal scheme falls into one of four types: Full Coverage, Partial Coverage, Immediate Revelation, and Non-Bonus. In a model of exponential bandit, we find that there is a dynamical switch of the types along the learning trajectory. Although learning stops efficiently, information is revealed too slowly compared with the planner's optimal solution.

econ.TH↗

arXiv2022

An Investigation into Whitening Loss for Self-supervised Learning

A desirable objective in self-supervised learning (SSL) is to avoid feature collapse. Whitening loss guarantees collapse avoidance by minimizing the distance between embeddings of positive pairs under the conditioning that the embeddings from different views are whitened. In this paper, we propose a framework with an informative indicator to analyze whitening loss, which provides a clue to demystify several interesting phenomena as well as a pivoting point connecting to other SSL methods. We reveal that batch whitening (BW) based methods do not impose whitening constraints on the embedding, but they only require the embedding to be full-rank. This full-rank constraint is also sufficient to avoid dimensional collapse. Based on our analysis, we propose channel whitening with random group partition (CW-RGP), which exploits the advantages of BW-based methods in preventing collapse and avoids their disadvantages requiring large batch size. Experimental results on ImageNet classification and COCO object detection reveal that the proposed CW-RGP possesses a promising potential for learning good representations. The code is available at https://github.com/winci-ai/CW-RGP.

cs.CV↗

arXiv2022

Stage-Aware Feature Alignment Network for Real-Time Semantic Segmentation of Street Scenes

Over the past few years, deep convolutional neural network-based methods have made great progress in semantic segmentation of street scenes. Some recent methods align feature maps to alleviate the semantic gap between them and achieve high segmentation accuracy. However, they usually adopt the feature alignment modules with the same network configuration in the decoder and thus ignore the different roles of stages of the decoder during feature aggregation, leading to a complex decoder structure. Such a manner greatly affects the inference speed. In this paper, we present a novel Stage-aware Feature Alignment Network (SFANet) based on the encoder-decoder structure for real-time semantic segmentation of street scenes. Specifically, a Stage-aware Feature Alignment module (SFA) is proposed to align and aggregate two adjacent levels of feature maps effectively. In the SFA, by taking into account the unique role of each stage in the decoder, a novel stage-aware Feature Enhancement Block (FEB) is designed to enhance spatial details and contextual information of feature maps from the encoder. In this way, we are able to address the misalignment problem with a very simple and efficient multi-branch decoder structure. Moreover, an auxiliary training strategy is developed to explicitly alleviate the multi-scale object problem without bringing additional computational costs during the inference phase. Experimental results show that the proposed SFANet exhibits a good balance between accuracy and speed for real-time semantic segmentation of street scenes. In particular, based on ResNet-18, SFANet respectively obtains 78.1% and 74.7% mean of class-wise Intersection-over-Union (mIoU) at inference speeds of 37 FPS and 96 FPS on the challenging Cityscapes and CamVid test datasets by using only a single GTX 1080Ti GPU.

cs.CV↗

arXiv2022

Deep Multi-Branch Aggregation Network for Real-Time Semantic Segmentation in Street Scenes

Real-time semantic segmentation, which aims to achieve high segmentation accuracy at real-time inference speed, has received substantial attention over the past few years. However, many state-of-the-art real-time semantic segmentation methods tend to sacrifice some spatial details or contextual information for fast inference, thus leading to degradation in segmentation quality. In this paper, we propose a novel Deep Multi-branch Aggregation Network (called DMA-Net) based on the encoder-decoder structure to perform real-time semantic segmentation in street scenes. Specifically, we first adopt ResNet-18 as the encoder to efficiently generate various levels of feature maps from different stages of convolutions. Then, we develop a Multi-branch Aggregation Network (MAN) as the decoder to effectively aggregate different levels of feature maps and capture the multi-scale information. In MAN, a lattice enhanced residual block is designed to enhance feature representations of the network by taking advantage of the lattice structure. Meanwhile, a feature transformation block is introduced to explicitly transform the feature map from the neighboring branch before feature aggregation. Moreover, a global context block is used to exploit the global contextual information. These key components are tightly combined and jointly optimized in a unified network. Extensive experimental results on the challenging Cityscapes and CamVid datasets demonstrate that our proposed DMA-Net respectively obtains 77.0% and 73.6% mean Intersection over Union (mIoU) at the inference speed of 46.7 FPS and 119.8 FPS by only using a single NVIDIA GTX 1080Ti GPU. This shows that DMA-Net provides a good tradeoff between segmentation quality and speed for semantic segmentation in street scenes.

cs.CV↗

arXiv2020

A VCG-based Fair Incentive Mechanism for Federated Learning

The enduring value of the Vickrey-Clarke-Groves (VCG) mechanism has been highlighted due to its adoption by Facebook ad auctions. Our research delves into its utility in the collaborative virtual goods production (CVGP) game, which finds application in realms like federated learning and crowdsourcing, in which bidders take on the roles of suppliers rather than consumers. We introduce the Procurement-VCG (PVCG) sharing rule into existing VCG mechanisms such that they can handle capacity limits and the continuous strategy space characteristic of the reverse auction setting in CVGP games. Our main theoretical contribution provides mathematical proofs to show that PVCG is the first in the CVGP game context to simultaneously achieve truthfulness, Pareto efficiency, individual rationality, and weak budget balance. These properties suggest the potential for Pareto-efficient production in the digital planned economy. Moreover, to compute the PVCG payments in a noisy economic environment, we propose the Report-Interpolation-Maximization (RIM) method. RIM facilitates the learning of the optimal procurement level and PVCG payments through iterative interactions with suppliers.

cs.GT↗

arXiv2020

Optimal Procurement Auction for Cooperative Production of Virtual Products: Vickrey-Clarke-Groves Meet Cremer-McLean

We set up a supply-side game-theoretic model for the cooperative production of virtual products. In our model, a group of producers collaboratively produce a virtual product by contributing costly input resources to a production coalition. Producers are capacitated, i.e., they cannot contribute more resources than their capacity limits. Our model is an abstraction of emerging internet-based business models such as federated learning and crowd computing. To maintain an efficient and stable production coalition, the coordinator should share with producers the income brought by the virtual product. Besides the demand-side information asymmetry, another two sources of supply-side information asymmetry intertwined in this problem: 1) the capacity limit of each producer and 2) the cost incurred to each producer. In this paper, we rigorously prove that a supply-side mechanism from the VCG family, PVCG, can overcome such multiple information asymmetry and guarantee truthfulness. Furthermore, with some reasonable assumptions, PVCG simultaneously attains truthfulness, ex-post allocative efficiency, ex-post individual rationality, and ex-post weak budget balancedness on the supply side, easing the well-known tension between these four objectives in the mechanism design literature.

cs.GT↗