arXiv subjects

Xiaoyi Lu

Publications and source records attributed to Xiaoyi Lu.

18 recordsLinked to original sources

arXiv2026

Contrastive Conditional-Unconditional Alignment for Long-tailed Diffusion Model

Training data for class-conditional image synthesis often exhibit a long-tailed distribution with limited amount of images for tail classes. Such an imbalance causes mode collapse and reduces the diversity of synthesized images for tail classes. For class-conditional diffusion models trained with imbalanced data, we aim to improve the diversity and fidelity of tail class images without compromising the quality of head class images. We propose contrastive conditional-unconditional alignment (CCUA), which comprises two synergistic loss functions. Our first loss is an Alignment Loss (AL) that aligns class-conditional generation with unconditional generation at large timesteps. Alignment loss makes the denoising process insensitive to class conditions for the initial steps, which enriches tail classes through knowledge sharing from head classes. Secondly, we diversify unconditional generation via an Unsupervised Contrastive Loss (UCL) to increase the distance/dissimilarity among synthetic images. We combine the two losses to implicitly diversify conditional generation. Our framework is easy to implement as demonstrated on both U-Net based architecture and Diffusion Transformer. Our method outperforms vanilla denoising diffusion probabilistic models, score-based diffusion model, and alternative contrastive methods for class-imbalanced image generation across various datasets, in particular ImageNet-LT with 256$\times$256 resolution.

cs.CV↗

arXiv2026

SPARe: Stacked Parallelism with Adaptive Reordering for Fault-Tolerant LLM Pretraining Systems with 100k+ GPUs

In large-scale LLM pre-training systems with 100k+ GPUs, failures become the norm rather than the exception, and restart costs can dominate wall-clock training time. However, existing fault-tolerance mechanisms are largely unprepared for this restart-dominant regime. To address this challenge, we propose SPARe - Stacked Parallelism with Adaptive Reordering - a fault-tolerance framework that masks node failures during gradient synchronization by stacking redundant data shards across parallelism groups and adaptively reordering execution. SPARe achieves availability comparable to traditional replication while maintaining near-constant computation overhead of only 2~3x, even under high redundancy where traditional replication would require linearly inflating overhead. We derive closed-form expressions for endurable failure count and computation overhead, validate them via SimGrid-based discrete-event simulation, and jointly optimize redundancy and checkpointing to minimize time-to-train. At extreme scale with up to 600k GPUs, SPARe reduces time-to-train by 40~50% compared to traditional replication.

cs.DC↗

arXiv2026

SplitFT: An Adaptive Federated Split Learning System For LLMs Fine-Tuning

Federated Split Learning has been identified as an efficient approach to address the computational resource constraints of clients in classical federated learning, while guaranteeing data privacy for distributed model training across data owners. However, it faces some critical challenges when such a training strategy meets large language models (LLMs) for fine-tuning. Such challenges include setting the cutlayer adaptively across different clients to address the data and device heterogeneity issues, which affect the system performance significantly. In addition, efficiently reducing the communication overhead during the fine-tuning procedure is also another challenge. No work tries to address these challenges. To bridge this gap, we propose SplitTF, an adaptive federated split learning system for LLMs fine-tuning. SplitFT enables different clients to set different cut layers according to their computation resources and trained model performance. SplitFT also proposes to reduce the LoRA rank in cutlayer to reduce the communication overhead. In addition to simulating the heterogeneous data in real-world applications for our proposed split federated learning system, we propose a length-based Dirichlet approach to divide the training data into different clients. Extensive experimental results show that our proposed approach outperforms the state-of-the-art approach for fine-tuning time efficiency and model performance based on various popular benchmarks.

cs.DC↗

arXiv2026

Nested Fourier-enhanced neural operator for efficient modeling of radiation transfer in fires

Computational fluid dynamics (CFD) has become an essential tool for predicting fire behavior, yet maintaining both efficiency and accuracy remains challenging. A major source of computational cost in fire simulations is the modeling of radiation transfer, which is usually the dominant heat transfer mechanism in fires. Solving the high-dimensional radiative transfer equation (RTE) with traditional numerical methods can be a performance bottleneck. Here, we present a machine learning framework based on Fourier-enhanced multiple-input neural operators (Fourier-MIONet) as an efficient alternative to direct numerical integration of the RTE. We first investigate the performance of neural operator architectures for a small-scale 2D pool fire and find that Fourier-MIONet provides the most accurate radiative solution predictions. The approach is then extended to 3D CFD fire simulations, where the computational mesh is locally refined across multiple levels. In these high-resolution settings, monolithic surrogate models for direct field-to-field mapping become difficult to train and computationally inefficient. To address this issue, a nested Fourier-MIONet is proposed to predict radiation solutions across multiple mesh-refinement levels. We validate the approach on 3D McCaffrey pool fires simulated with FireFOAM, including fixed fire sizes and a unified model trained over a continuous range of heat release rates (HRRs). The proposed method achieves global relative errors of 2-4% for 3D varying-HRR scenarios while providing faster inference than the estimated cost of one finite-volume radiation solve in FireFOAM for the 16-solid-angle case. With fast and accurate inference, the surrogate makes higher-fidelity radiation treatments practical and enables the incorporation of more spectrally resolved radiation models into CFD fire simulations for engineering applications.

physics.flu-dyn↗

arXiv2026

Architectural Foundations for Checkpointing and Restoration in Quantum HPC Systems

In this work, we explore the design of the checkpointing and restoration for quantum HPC that leverages dynamic circuit technology to enable restartable and resilient quantum execution. Rather than attempting to checkpoint quantum states, our approach redefines checkpointing as a control flow and algorithmic state problem. By exploiting mid-circuit measurements, classical feed forward, and conditional execution supported by dynamic circuits, we capture sufficient program state to allow correct restoration of quantum workflows after interruption or failure. This design aligns naturally with iterative and staged quantum algorithms such as variational eigensolvers, quantum approximate optimization, and time-stepping methods commonly used in quantum simulation and scientific computing.

quant-ph↗

arXiv2026

Robust Ultra-High-Dimensional Variable Selection With Correlated Structure Using Group Testing

Background: High-dimensional genomic data exhibit strong group correlation structures that challenge conventional feature selection methods, which often assume feature independence or rely on pre-defined pathways and are sensitive to outliers and model misspecification. Methods: We propose the Dorfman screening framework, a multi-stage procedure that forms data-driven variable groups via hierarchical clustering, performs group and within-group hypothesis testing, and refines selection using elastic net or adaptive elastic net. Robust variants incorporate OGK-based covariance estimation, rank-based correlation, and Huber-weighted regression to handle contaminated and non-normal data. Results: In simulations, Dorfman-Sparse-Adaptive-EN performed best under normal conditions, while Robust-OGK-Dorfman-Adaptive-EN showed clear advantages under data contamination, outperforming classical Dorfman and competing methods. Applied to NSCLC gene expression data for trametinib response, robust Dorfman methods achieved the lowest prediction errors and enriched recovery of clinically relevant genes. Conclusions: The Dorfman framework provides an efficient and robust approach to genomic feature selection. Robust-OGK-Dorfman-Adaptive-EN offers strong performance under both ideal and contaminated conditions and scales to ultra-high-dimensional settings, making it well suited for modern genomic biomarker discovery.

cs.LG↗

arXiv2026

Benchmarking neural surrogates on realistic spatiotemporal multiphysics flows

Predicting multiphysics dynamics is computationally expensive and challenging due to the severe coupling of multi-scale, heterogeneous physical processes. While neural surrogates promise a paradigm shift, the field currently suffers from an "illusion of mastery", as repeatedly emphasized in top-tier commentaries: existing evaluations overly rely on simplified, low-dimensional proxies, which fail to expose the models' inherent fragility in realistic regimes. To bridge this critical gap, we present REALM (REalistic AI Learning for Multiphysics), a rigorous benchmarking framework designed to test neural surrogates on challenging, application-driven reactive flows. REALM features 11 high-fidelity datasets spanning from canonical multiphysics problems to complex propulsion and fire safety scenarios, alongside a standardized end-to-end training and evaluation protocol that incorporates multiphysics-aware preprocessing and a robust rollout strategy. Using this framework, we systematically benchmark over a dozen representative surrogate model families, including spectral operators, convolutional models, Transformers, pointwise operators, and graph/mesh networks, and identify three robust trends: (i) a scaling barrier governed jointly by dimensionality, stiffness, and mesh irregularity, leading to rapidly growing rollout errors; (ii) performance primarily controlled by architectural inductive biases rather than parameter count; and (iii) a persistent gap between nominal accuracy metrics and physically trustworthy behavior, where models with high correlations still miss key transient structures and integral quantities. Taken together, REALM exposes the limits of current neural surrogates on realistic multiphysics flows and offers a rigorous testbed to drive the development of next-generation physics-aware architectures.

cs.LG↗

arXiv2026

Learning Provably Correct Distributed Protocols Without Human Knowledge

Provably correct distributed protocols, which are a critical component of modern distributed systems, are highly challenging to design and have often required decades of human effort. These protocols allow multiple agents to coordinate to come to a common agreement in an environment with uncertainty and failures. We formulate protocol design as a search problem over strategies in a game with imperfect information, and the desired correctness conditions are specified in Satisfiability Modulo Theories (SMT). However, standard methods for solving multi-agent games fail to learn correct protocols in this setting, even when the number of agents is small. We propose a learning framework, GGMS, which integrates a specialized variant of Monte Carlo Tree Search with a transformer-based action encoder, a global depth-first search to break out of local minima, and repeated feedback from a model checker. Protocols output by GGMS are verified correct via exhaustive model checking for all executions within the bounded setting. We further prove that, under mild assumptions, the search process is complete: if a correct protocol exists, GGMS will eventually find it. In experiments, we show that GGMS can learn correct protocols for larger settings than existing methods.

cs.AI↗

arXiv2025

MoE-Compression: How the Compression Error of Experts Affects the Inference Accuracy of MoE Model?

With the widespread application of Mixture of Experts (MoE) reasoning models in the field of LLM learning, efficiently serving MoE models under limited GPU memory constraints has emerged as a significant challenge. Offloading the non-activated experts to main memory has been identified as an efficient approach to address such a problem, while it brings the challenges of transferring the expert between the GPU memory and main memory. We need to explore an efficient approach to compress the expert and analyze how the compression error affects the inference performance. To bridge this gap, we propose employing error-bounded lossy compression algorithms (such as SZ3 and CuSZp) to compress non-activated experts, thereby reducing data transfer overhead during MoE inference. We conduct extensive experiments across various benchmarks and present a comprehensive analysis of how compression-induced errors in different experts affect overall inference accuracy. The results indicate that experts in the shallow layers, which are primarily responsible for the attention mechanism and the transformation of input tokens into vector representations, exhibit minimal degradation in inference accuracy when subjected to bounded errors. In contrast, errors in the middle-layer experts, which are central to model reasoning, significantly impair inference accuracy. Interestingly, introducing bounded errors in the deep-layer experts, which are mainly responsible for instruction following and output integration, can sometimes lead to improvements in inference accuracy.

cs.LG↗

arXiv2025

ZCCL: Significantly Improving Collective Communication With Error-Bounded Lossy Compression

With the ever-increasing computing power of supercomputers and the growing scale of scientific applications, the efficiency of MPI collective communication turns out to be a critical bottleneck in large-scale distributed and parallel processing. The large message size in MPI collectives is particularly concerning because it can significantly degrade overall parallel performance. To address this issue, prior research simply applies off-the-shelf fixed-rate lossy compressors in the MPI collectives, leading to suboptimal performance, limited generalizability, and unbounded errors. In this paper, we propose a novel solution, called ZCCL, which leverages error-bounded lossy compression to significantly reduce the message size, resulting in a substantial reduction in communication costs. The key contributions are three-fold. (1) We develop two general, optimized lossy-compression-based frameworks for both types of MPI collectives (collective data movement as well as collective computation), based on their particular characteristics. Our framework not only reduces communication costs but also preserves data accuracy. (2) We customize fZ-light, an ultra-fast error-bounded lossy compressor, to meet the specific needs of collective communication. (3) We integrate ZCCL into multiple collectives, such as Allgather, Allreduce, Scatter, and Broadcast, and perform a comprehensive evaluation based on real-world scientific application datasets. Experiments show that our solution outperforms the original MPI collectives as well as multiple baselines by 1.9--8.9X.

cs.DC↗

arXiv2024

Ba-ZebraConf: A Three-Dimension Bayesian Framework for Efficient System Troubleshooting

The proliferation of heterogeneous configurations in distributed systems presents significant challenges in ensuring stability and efficiency. Misconfigurations, driven by complex parameter interdependencies, can lead to critical failures. Group Testing (GT) has been leveraged to expedite troubleshooting by reducing the number of tests, as demonstrated by methods like ZebraConf. However, ZebraConf's binary-splitting strategy suffers from sequential testing, limited handling of parameter interdependencies, and susceptibility to errors such as noise and dilution. We propose Ba-ZebraConf, a novel three-dimensional Bayesian framework that addresses these limitations. It integrates (1) Bayesian Group Testing (BGT), which employs probabilistic lattice models and the Bayesian Halving Algorithm (BHA) to dynamically refine testing strategies, prioritizing high-informative parameters and adapting to real-time outcomes. Bayesian optimization tunes hyperparameters, such as pool sizes and test thresholds, to maximize testing efficiency. (2) Bayesian Optimization (BO) to automate hyperparameter tuning for test efficiency, and (3) Bayesian Risk Refinement (BRR) to iteratively capture parameter interdependencies and improve classification accuracy. Ba-ZebraConf adapts to noisy environments, captures parameter interdependencies, and scales effectively for large configuration spaces. Experimental results show that Ba-ZebraConf reduces test counts and execution time by 67% compared to ZebraConf while achieving 0% false positives and false negatives. These results establish Ba-ZebraConf as a robust and scalable solution for troubleshooting heterogeneous distributed systems.

eess.SY↗

arXiv2024

gZCCL: Compression-Accelerated Collective Communication Framework for GPU Clusters

GPU-aware collective communication has become a major bottleneck for modern computing platforms as GPU computing power rapidly rises. A traditional approach is to directly integrate lossy compression into GPU-aware collectives, which can lead to serious performance issues such as underutilized GPU devices and uncontrolled data distortion. In order to address these issues, in this paper, we propose gZCCL, a first-ever general framework that designs and optimizes GPU-aware, compression-enabled collectives with an accuracy-aware design to control error propagation. To validate our framework, we evaluate the performance on up to 512 NVIDIA A100 GPUs with real-world applications and datasets. Experimental results demonstrate that our gZCCL-accelerated collectives, including both collective computation (Allreduce) and collective data movement (Scatter), can outperform NCCL as well as Cray MPI by up to 4.5X and 28.7X, respectively. Furthermore, our accuracy evaluation with an image-stacking application confirms the high reconstructed data quality of our accuracy-aware framework.

cs.DC↗

arXiv2024

An Optimized Error-controlled MPI Collective Framework Integrated with Lossy Compression

With the ever-increasing computing power of supercomputers and the growing scale of scientific applications, the efficiency of MPI collective communications turns out to be a critical bottleneck in large-scale distributed and parallel processing. The large message size in MPI collectives is particularly concerning because it can significantly degrade the overall parallel performance. To address this issue, prior research simply applies the off-the-shelf fix-rate lossy compressors in the MPI collectives, leading to suboptimal performance, limited generalizability, and unbounded errors. In this paper, we propose a novel solution, called C-Coll, which leverages error-bounded lossy compression to significantly reduce the message size, resulting in a substantial reduction in communication cost. The key contributions are three-fold. (1) We develop two general, optimized lossy-compression-based frameworks for both types of MPI collectives (collective data movement as well as collective computation), based on their particular characteristics. Our framework not only reduces communication cost but also preserves data accuracy. (2) We customize SZx, an ultra-fast error-bounded lossy compressor, to meet the specific needs of collective communication. (3) We integrate C-Coll into multiple collectives, such as MPI_Allreduce, MPI_Scatter, and MPI_Bcast, and perform a comprehensive evaluation based on real-world scientific datasets. Experiments show that our solution outperforms the original MPI collectives as well as multiple baselines and related efforts by 1.8-2.7X.

cs.DC↗

arXiv2022

Arcadia: A Fast and Reliable Persistent Memory Replicated Log

The performance properties of byte-addressable persistent memory (PMEM) have the potential to significantly improve system performance over a wide spectrum of applications. But persistent memory brings considerable new challenges to the programmer: only 8-byte write atomicity, out of order flush and availability limited by node failure. It's possible to work with the atomicity and ordering constraints of PMEM directly by carefully sequencing the order of store operations and inserting explicit flush and fence operations at each ordering point. But this is tedious and error-prone: too many flush operations defeat the performance benefits of PMEM, and even with generous use, it is difficult to prove that a given program is crash-consistent. Logging is a great abstraction to deal with these issues but prior work on PMEM logging has not successfully hidden the idiosyncrasies of PMEM. Moreover, shortcomings in the log interface and design have prevented attainment of full PMEM performance. We believe that a log design that hides the idiosyncrasies from programmers while delivering full performance is key to success. In this paper, we present the design and implementation of Arcadia, a generic replicated log on PMEM to address these problems. Arcadia handles atomicity, integrity, and replication of log records to reduce programmer burden. Our design has several novel aspects including concurrent log writes with in-order commit, atomicity and integrity primitives for local and remote PMEM writes, and a frequency-based log force policy for providing low overhead persistence with guaranteed bounded loss of uncommitted records. Our evaluation shows that Arcadia outperforms state-of-the-art PMEM logs, such as PMDK's libpmemlog, FLEX, and Query Fresh by several times while providing stronger log record durability guarantees. We expect Arcadia to become the leading off-the-shelf PMEM log design.

cs.DC↗

arXiv2021

Towards Offloadable and Migratable Microservices on Disaggregated Architectures: Vision, Challenges, and Research Roadmap

Microservice and serverless computing systems open up massive versatility and opportunity to distributed and datacenter-scale computing. In the meantime, the deployments of modern datacenter resources are moving to disaggregated architectures. With the flourishing growths from both sides, we believe this is high time to write this vision paper to propose a potential research agenda to achieve efficient deployments, management, and executions of next-generation microservices on top of the emerging disaggregated datacenter architectures. In particular, we envision a critical systems research direction of designing and developing offloadable and migratable microservices on disaggregated architectures. With this vision, we have surveyed the recent related work to demonstrate the importance and necessity of researching it. We also outline the fundamental challenges that distributed systems and datacenter-scale computing research may encounter. We further propose a research roadmap to achieve our envisioned objectives in a promising way. Within the roadmap, we identify potential techniques and methods that can be leveraged.

cs.DC↗

arXiv2018

Designing a Micro-Benchmark Suite to Evaluate gRPC for TensorFlow: Early Experiences

Remote procedure call (RPC) is the backbone of many modern distributed systems. Google's gRPC is one of the most popular open source RPC frameworks available in the community. gRPC is the main communication engine for Google's Deep Learning framework TensorFlow. TensorFlow primarily uses gRPC for communicating tensors and administrative tasks among different processes. Tensor updates during the training phase are communication intensive and thus TensorFlow's performance is heavily dependent on the underlying network and the efficacy of the communication engine. Training deep learning models on TensorFlow can take significant time ranging from several minutes to several hours, even several days. Thus system researchers need to devote a lot of time to understand the impact of communication on the overall performance. Clearly, there is lack of benchmarks available for system researchers. Therefore, we propose TF-gRPC-Bench micro-benchmark suite that enables system researchers to quickly understand the impact of the underlying network and communication runtime on deep learning workloads. To achieve this, we first analyze the characteristics of TensorFlow workload over gRPC by training popular deep learning models. Then, we propose three micro-benchmarks that take account these workload characteristics. In addition, we comprehensively evaluate gRPC with TF-gRPC-Bench micro-benchmark suite on different clusters over Ethernet, IPoIB, and RDMA, and present the results.

cs.DC↗

arXiv2014

Performance Benefits of DataMPI: A Case Study with BigDataBench

Apache Hadoop and Spark are gaining prominence in Big Data processing and analytics. Both of them are widely deployed on Internet companies. On the other hand, high-performance data analysis requirements are causing academical and industrial communities to adopt state-of-the-art technologies in HPC to solve Big Data problems. Recently, we have proposed a key-value pair based communication library, DataMPI, which is extending MPI to support Hadoop/Spark-like Big Data Computing jobs. In this paper, we use BigDataBench, a Big Data benchmark suite, to do comprehensive studies on performance and resource utilization characterizations of Hadoop, Spark and DataMPI. From our experiments, we observe that the job execution time of DataMPI has up to 55% and 39% speedups compared with those of Hadoop and Spark, respectively. Most of the benefits come from the high-efficiency communication mechanisms in DataMPI. We also notice that the resource (CPU, memory, disk and network I/O) utilizations of DataMPI are also more efficient than those of the other two frameworks.

cs.PF↗

arXiv2014

On Big Data Benchmarking

Big data systems address the challenges of capturing, storing, managing, analyzing, and visualizing big data. Within this context, developing benchmarks to evaluate and compare big data systems has become an active topic for both research and industry communities. To date, most of the state-of-the-art big data benchmarks are designed for specific types of systems. Based on our experience, however, we argue that considering the complexity, diversity, and rapid evolution of big data systems, for the sake of fairness, big data benchmarks must include diversity of data and workloads. Given this motivation, in this paper, we first propose the key requirements and challenges in developing big data benchmarks from the perspectives of generating data with 4V properties (i.e. volume, velocity, variety and veracity) of big data, as well as generating tests with comprehensive workloads for big data systems. We then present the methodology on big data benchmarking designed to address these challenges. Next, the state-of-the-art are summarized and compared, following by our vision for future research directions.

cs.PF↗