arXiv subjects

Mingyuan Xia

Publications and source records attributed to Mingyuan Xia.

7 recordsLinked to original sources

arXiv2026

HBM Is Not All You Need: Efficient Disaggregated LLM Serving across Memory-heterogeneous Accelerators

LLM inference comprises a compute-bound prefill phase and a memory-bound decode phase, and recent systems disaggregate them onto separate hardware. Yet today's datacenter GPUs rely on costly HBM whose bandwidth sits almost entirely idle during prefill. LLM serving across memory-heterogeneous accelerators (MemHA) pairs GDDR-based accelerators for prefill with HBM-based GPUs for decode, promising lower cost without sacrificing performance. Pushed to its most economical form, MemHA serving is inherently cross-vendor, since the best-suited chip for each phase may come from a different vendor. This breaks two assumptions that single-vendor disaggregation takes for granted -- a KV format both ends consume natively, and a shared software stack. We present \textbf{HMA-Serve}, a MemHA-centric disaggregated serving system pairing GDDR-based accelerators for prefill with HBM-based GPUs for decode efficiently. HMA-Serve achieves this through (1) phase-wise quantization, applying vendor-native low precision for high-throughput prefill while keeping decode in high-precision BF16, (2) a compute-transfer pipeline that overlaps each layer's KV cache transfer with later-layer prefill to reduce time-to-first-token (TTFT), and (3) deferred dequantization, shipping raw quantized bytes and reconstructing them lazily on the decode GPU to reduce network bandwidth and HBM usage. Across four Qwen3 models (4B--32B) and three production traces, HMA-Serve delivers up to $3.2\times$ higher goodput than state-of-the-art memory-homogeneous methods and $4.8\times$ higher goodput-per-dollar, with no measurable loss on generation-quality benchmarks.

cs.AR↗

arXiv2026

The Phantom of PCIe: Constraining Generative Artificial Intelligences for Practical Peripherals Trace Synthesizing

Peripheral Component Interconnect Express (PCIe) is the de facto interconnect standard for high-speed peripherals and CPUs. The development of PCIe devices for emerging applications requires realistic Transaction Layer Packet (TLP) traces that accurately simulate device-CPU interactions. While generative AI offers a promising avenue for synthesizing complex TLP sequences, it is prone to a critical challenge inherent in all generation tasks: hallucination. Naively applying these models often produces traces that violate fundamental PCIe protocol rules, such as ordering and causality, rendering them unusable for device simulation. To resolve this, our work introduces a methodology to bridge the gap between generative AI and high-fidelity device simulation. This paper presents Phantom, a framework that systematically addresses AI-generated hallucinations in TLP synthesis. Phantom achieves this by coupling a generative backbone with a novel post-processing filter that enforces PCIe-specific constraints, effectively eliminating invalid TLP sequences. We validate Phantom's effectiveness by synthesizing TLP traces for an actual PCIe network interface card. Experimental results show that Phantom produces practical, large-scale TLP traces, significantly outperforming existing models, with improvements of up to 1000$\times$ in task-specific metrics and up to 2.19$\times$ in Fréchet Inception Distance (FID) compared to backbone-only methods. The prototype implementation has been made open-source.

cs.LG↗

arXiv2026

Synecdoche: Efficient and Accurate In-Network Traffic Classification via Direct Packet Sequential Pattern Matching

Traffic classification on programmable data plane holds great promise for line-rate processing, with methods evolving from per-packet to flow-level analysis for higher accuracy. However, a trade-off between accuracy and efficiency persists. Statistical feature-based methods align with hardware constraints but often exhibit limited accuracy, while online deep learning methods using packet sequential features achieve superior accuracy but require substantial computational resources. This paper presents Synecdoche, the first traffic classification framework that successfully deploys packet sequential features on a programmable data plane via pattern matching, achieving both high accuracy and efficiency. Our key insight is that discriminative information concentrates in short sub-sequences--termed Key Segments--that serve as compact traffic features for efficient data plane matching. Synecdoche employs an "offline discovery, online matching" paradigm: deep learning models automatically discover Key Segment patterns offline, which are then compiled into optimized table entries for direct data plane matching. Extensive experiments demonstrate Synecdoche's superior accuracy, improving F1-scores by up to 26.4% against statistical methods and 18.3% against online deep learning methods, while reducing latency by 13.0% and achieving 79.2% reduction in SRAM usage. The source code of Synecdoche is publicly available to facilitate reproducibility and further research.

cs.NI↗

arXiv2025

TimeEmb: A Lightweight Static-Dynamic Disentanglement Framework for Time Series Forecasting

Temporal non-stationarity, the phenomenon that time series distributions change over time, poses fundamental challenges to reliable time series forecasting. Intuitively, the complex time series can be decomposed into two factors, \ie time-invariant and time-varying components, which indicate static and dynamic patterns, respectively. Nonetheless, existing methods often conflate the time-varying and time-invariant components, and jointly learn the combined long-term patterns and short-term fluctuations, leading to suboptimal performance facing distribution shifts. To address this issue, we initiatively propose a lightweight static-dynamic decomposition framework, TimeEmb, for time series forecasting. TimeEmb innovatively separates time series into two complementary components: (1) time-invariant component, captured by a novel global embedding module that learns persistent representations across time series, and (2) time-varying component, processed by an efficient frequency-domain filtering mechanism inspired by full-spectrum analysis in signal processing. Experiments on real-world datasets demonstrate that TimeEmb outperforms state-of-the-art baselines and requires fewer computational resources. We conduct comprehensive quantitative and qualitative analyses to verify the efficacy of static-dynamic disentanglement. This lightweight framework can also improve existing time-series forecasting methods with simple integration. To ease reproducibility, the code is available at https://github.com/showmeon/TimeEmb.

cs.LG↗

arXiv2025

AdaSVD: Adaptive Singular Value Decomposition for Large Language Models

Large language models (LLMs) have achieved remarkable success in natural language processing (NLP) tasks, yet their substantial memory requirements present significant challenges for deployment on resource-constrained devices. Singular Value Decomposition (SVD) has emerged as a promising compression technique for LLMs, offering considerable reductions in memory overhead. However, existing SVD-based methods often struggle to effectively mitigate the errors introduced by SVD truncation, leading to a noticeable performance gap when compared to the original models. Furthermore, applying a uniform compression ratio across all transformer layers fails to account for the varying importance of different layers. To address these challenges, we propose AdaSVD, an adaptive SVD-based LLM compression approach. Specifically, AdaSVD introduces adaComp, which adaptively compensates for SVD truncation errors by alternately updating the singular matrices $\mathcal{U}$ and $\mathcal{V}^\top$. Additionally, AdaSVD introduces adaCR, which adaptively assigns layer-specific compression ratios based on the relative importance of each layer. Extensive experiments across multiple LLM/VLM families and evaluation metrics demonstrate that AdaSVD consistently outperforms state-of-the-art (SOTA) SVD-based methods, achieving superior performance with significantly reduced memory requirements. Code and models of AdaSVD will be available at https://github.com/ZHITENGLI/AdaSVD.

cs.CV↗

arXiv2025

Equinox: Holistic Fair Scheduling in Serving Large Language Models

We address the limitations of current LLM serving with a dual-counter framework separating user and operator perspectives. The User Fairness Counter measures quality of service via weighted tokens and latency; the Resource Fairness Counter measures operational efficiency through throughput and GPU utilization. Since these metrics are only available post-execution, creating a scheduling paradox, we introduce a deterministic Mixture of Prediction Experts (MoPE) framework to predict user-perceived latency, output tokens, throughput, and GPU utilization. These predictions enable calculation of a unified Holistic Fairness score that balances both counters through tunable parameters for proactive fairness-aware scheduling. We implement this in Equinox, an open-source system with other optimizations like adaptive batching, and stall-free scheduling. Evaluations on production traces (ShareGPT, LMSYS) and synthetic workloads demonstrate Equinox achieves up to $1.3\times$ higher throughput, 60\% lower time-to-first-token latency, and 13\% higher fairness versus VTC while maintaining 94\% GPU utilization, proving fairness under bounded discrepancy across heterogeneous platforms.

cs.DC↗

arXiv2015

Does "Like" Really Mean Like? A Study of the Facebook Fake Like Phenomenon and an Efficient Countermeasure

Social networks help to bond people who share similar interests all over the world. As a complement, the Facebook "Like" button is an efficient tool that bonds people with the online information. People click on the "Like" button to express their fondness of a particular piece of information and in turn tend to visit webpages with high "Like" count. The important fact of the Like count is that it reflects the number of actual users who "liked" this information. However, according to our study, one can easily exploit the defects of the "Like" button to counterfeit a high "Like" count. We provide a proof-of-concept implementation of these exploits, and manage to generate 100 fake Likes in 5 minutes with a single account. We also reveal existing counterfeiting techniques used by some online sellers to achieve unfair advantage for promoting their products. To address this fake Like problem, we study the varying patterns of Like count and propose an innovative fake Like detection method based on clustering. To evaluate the effectiveness of our algorithm, we collect the Like count history of more than 9,000 websites. Our experiments successfully uncover 16 suspicious fake Like buyers that show abnormal Like count increase patterns.

cs.SI↗