arXiv subjects

Zhiyuan Guo

Publications and source records attributed to Zhiyuan Guo.

At least 19 recordsLinked to original sources

arXiv2026

Battery-Swapping Station Operation Under Forecast Uncertainty: A Scenario-Based Stochastic MPC Framework

Battery-swapping stations (BSSs) can shorten electric-vehicle energy replenishment while using centrally managed battery inventories as flexible grid-connected storage. Realizing both benefits requires the station to schedule charging, grid discharge, and swapping service before future customer demand and electricity prices are known. This paper develops a forecast-aware rolling-horizon operating framework for this problem. A lightweight DLinear model predicts 24-hour price and demand trajectories, Stein variational gradient descent quantifies their uncertainty through representative scenarios, and a two-stage stochastic model predictive controller converts those scenarios into station decisions. The controller accounts for service shortfall, terminal readiness, a protected service buffer, and electrochemical degradation without assuming perfect future information. The application contribution is an implementable controller that coordinates the station's mobility-service and energy-storage roles. The methodological contribution is a modular forecast-to-control interface that separates the operational value of mean-forecast accuracy from that of uncertainty representation. In a 120-day closed-loop evaluation, DLinear-SVGD SMPC achieves the lowest cost among the implementable controllers. Relative to deterministic DLinear MPC, it reduces final cost by 1.2\% and service-shortfall hours by 80.7\%, with 99.10\% of the evaluated hours free of shortfall.

math.OC↗

arXiv2026

Focal-point scanning for dose delivery and optimization with focused laser-accelerated very-high-energy electron beams

Focused very-high-energy electron (VHEE) beams can produce localized dose enhancement at selected depths, but irradiation of a finite target requires coordinated control of multiple focal positions, incidence directions, and beam weights while limiting exposure of nearby organs at risk (OARs). We present Focal-Point Scanning (FPS), a dose delivery and optimization method developed for laser wakefield accelerator (LWFA)-driven VHEE beams. The method is based on a two-dipole focusing system that produces single-plane beam convergence and allows the focal position to be varied by changing the magnetic field strength. FPS distributes focal points throughout the planning target volume and determines focal-point-specific incidence sectors according to the geometry of nearby critical OARs. The method was evaluated using the AAPM TG119 C-shape benchmark and one previously treated lung radiotherapy case. At matched target coverage, FPS reduced the TG119 Core mean dose by approximately one half relative to parallel VHEE and intensity-modulated x-ray plans, approaching the single-field proton pencil-beam-scanning reference. In the lung case, FPS maintained target coverage comparable to the clinical volumetric modulated arc therapy reference while reducing the mean dose to every evaluated OAR; spinal-cord mean and maximum doses decreased by 93.2% and 87.2%, respectively. The evaluated OAR mean doses varied little across rms energy spreads of 0 to 10% and for a flat-top electron spectrum spanning 150 to 250 MeV. These results demonstrate that focal-point-specific angular selection can translate focused-beam physics into effective OAR sparing and support FPS as a planning strategy for broadband LWFA-VHEE radiotherapy.

physics.med-ph↗

arXiv2026

Absolute charge calibration of DRZ phosphor screens for relativistic electron bunches

Laser-plasma accelerators have been the subject of extensive research in recent years. The electron beams they generate exhibit a broad energy spread. To conveniently characterize beams from laser wakefield acceleration (LWFA), electron spectrometers employing scintillating screens coupled with CCD cameras are typically used. In this work, we calibrate a series of DRZ phosphor screens and measure the spectra of the light they emit. The calibration was performed using the radio-frequency linear electron accelerator at Tsinghua University, which provided monoenergetic electron beams with peak energy of approximately 30 MeV.

physics.acc-ph↗

arXiv2026

Mandol: An Agglomerative Agent Memory System for Long-Term Conversations

Long-term conversational agents need to remember and query cross-session, multi-typed information with complex correlations. Existing agent memory systems rely on heterogeneous vector and graph databases, which fragment memory information and cause high cross-database I/O latency. For retrieval, common RAG-style methods tend to introduce noise, miss correlated clues, and lack token budget control, degrading LLM accuracy and efficiency. We propose Mandol, an agglomerative memory system that consolidates fragmented memory representations and storage into a unified memory-native architecture. Its core components include: (1) a hierarchical memory model that organizes memory into a basic layer representing raw memory information and a high-level abstract layer that agglomerates basic memories into traceable abstract memories, both uniformly represented as structured semantic graphs; (2) an agglomerative semantic data structure combining SemanticMap and SemanticGraph, which natively fuses key-value, vector, and graph structures and provides unified hybrid retrieval operators to eliminate cross-database I/O; and (3) a quantitative query mechanism with query-adaptive routing, quantitative denoising and conflict resolution, and token-constrained context generation, all without involving LLMs during retrieval. Experiments on two widely used long-term conversation benchmarks, LoCoMo and LongMemEval, show that Mandol achieves the best overall accuracy among representative agent memory systems. For performance comparison, Mandol also obtains a 5.4x retrieval speedup and a 4.8x insertion speedup under 10 QPS concurrent load, while still maintaining low latency on consumer-grade hardware.

cs.DB↗

arXiv2026

VDCores: Resource Decoupled Programming and Execution for Asynchronous GPU

Modern GPUs increasingly rely on specialized and asynchronous hardware units to deliver high performance. Yet these units are often underutilized because today's GPU software stacks still organize programming and execution around a monolithic kernel model that mismatches asynchronous hardware. To address this issue, Virtual Decoupled Engines (VDCores) presents a new decoupled programming and execution model for asynchronous GPUs. VDCores abstracts asynchronous hardware execution units as resource isolated virtual cores and represents workloads as dependency-connected micro-operations (micro-ops). this abstraction removes static orchestration from the programmer, enables automatic overlap of memory and compute based on dependency and resource readiness, and thereby improves utilization of asynchronous hardware resources. Realizing such a decoupled abstraction efficiently on today's GPUs is itself challenging, VDCores addresses this through a GPU-specialized programming model and GPU runtime design that preserves the flexibility while minimizing implementation overhead. Across four LLM inference workloads on GH200, H100, and RTX 6000 Pro GPUs, VDCores significantly improves decoding throughput by 24% on average and by up to 77% under dynamic inputs, while reducing kernel programming and specialization effort by 90%. We have open sourced VDCores at https://github.com/vdcores/vdcores.

cs.DC↗

arXiv2026

DAK: Direct-Access-Enabled GPU Memory Offloading with Optimal Efficiency for LLM Inference

LLM inference is constrained by GPU memory capacity and bandwidth. Tiered memory architectures mitigate this by allowing the GPU to offload memory to the remote tier. However, existing memory offloading frameworks rely on prefetching data into local GPU HBM. This approach underutilizes system resources by introducing HBM contention, squandering memory capacity, and creating pipeline bubbles. We show that enabling direct GPU access to remote memory significantly outperforms prefetching, achieving optimal aggregate system bandwidth. We propose DAK, an end-to-end direct-access memory offloading framework that repurposes the Tensor Memory Accelerator (TMA) to asynchronously fetch offloaded weights and KV caches directly from remote memory into GPU shared memory (SMEM). To maximize remote access performance, DAK introduces a greedy algorithm to determine optimal per-operation offloading ratios, alongside active congestion control and TMA multicast to eliminate interconnect bottlenecks and read amplification. Evaluations across diverse architectures show that DAK achieves near-optimal bandwidth aggregation, with up to 3$\times$ performance gains on NVLink-C2C and 1.8$\times$ on PCIe systems compared to state-of-the-art memory offloading baselines.

cs.DC↗

arXiv2026

IRMaGiC: Extending Luminous Red Galaxy Selection into the Infrared with Joint Rubin Observatory's Large Survey of Space Time and Roman's High Latitude Imaging Survey

We introduce IRMaGiC, an algorithm built based on RedMaGiC desgined to enhance the selection of Luminous Red Galaxies (LRGs) across the redshift range $1 \leq z \leq 2$. We show that this method extends the capabilities of the redMaGiC algorithm by applying it to simulated photometric data from the Vera C. Rubin Observatory's Legacy Survey of Space and Time (LSST) and the Nancy Grace Roman Space Telescope's High Latitude Wide Area Survey (HLWAS). By integrating infrared band coverage from Roman HLWAS with LSST's optical bands, IRMaGiC enables red-sequence calibration at higher redshifts. We demonstrate that IRMaGiC reduces scatter and bias in photometric redshift estimates for LRGs at higher redshift, providing more accurate redshift assessments compared to existing methods. Our findings suggest that incorporating infrared data can considerably improve the selection and redshift estimation of LRGs at higher redshift, offering substantial benefits for future cosmological surveys.

astro-ph.CO↗

arXiv2025

Game-RL: Synthesizing Multimodal Verifiable Game Data to Boost VLMs' General Reasoning

Vision-language reinforcement learning (RL) has primarily focused on narrow domains (e.g. geometry or chart reasoning). This leaves broader training scenarios and resources underexplored, limiting the exploration and learning of Vision Language Models (VLMs) through RL. We find video games inherently provide rich visual elements and mechanics that are easy to verify. To fully use the multimodal and verifiable reward in video games, we propose Game-RL, constructing diverse game tasks for RL training to boost VLMs general reasoning ability. To obtain training data, we propose Code2Logic, a novel approach that adapts game code to synthesize game reasoning task data, thus obtaining the GameQA dataset of 30 games and 158 tasks with controllable difficulty gradation. Unexpectedly, RL training solely on GameQA enables multiple VLMs to achieve performance improvements across 7 diverse vision-language benchmarks, demonstrating the value of Game-RL for enhancing VLMs' general reasoning. Furthermore, this suggests that video games may serve as valuable scenarios and resources to boost general reasoning abilities. Our code, dataset and models are available at the GitHub repository.

cs.CL↗

arXiv2025

Learning Semantics, Not Addresses: Runtime Neural Prefetching for Far Memory

Memory prefetching has long boosted CPU caches and is increasingly vital for far-memory systems, where large portions of memory are offloaded to cheaper, remote tiers. While effective prefetching requires accurate prediction of future accesses, prior ML approaches have been limited to simulation or small-scale hardware. We introduce FarSight, the first Linux-based far-memory system to leverage deep learning by decoupling application semantics from runtime memory layout. This separation enables offline-trained models to predict access patterns over a compact ordinal vocabulary, which are resolved at runtime through lightweight mappings. Across four data-intensive workloads, FarSight delivers up to 3.6x higher performance than the state-of-the-art.

cs.LG↗

arXiv2025

An Early Exploration of Deep-Learning-Driven Prefetching for Far Memory

Far-memory systems, where applications store less-active data in more energy-efficient memory media, are increasingly adopted by data centers. However, applications are bottlenecked by on-demand data fetching from far- to local-memory. We present Memix, a far-memory system that embodies a deep-learning-system co-design for efficient and accurate prefetching, minimizing on-demand far-memory accesses. One key observation is that memory accesses are shaped by both application semantics and runtime context, providing an opportunity to optimize each independently. Preliminary evaluation of Memix on data-intensive workloads shows that it outperforms the state-of-the-art far-memory system by up to 42%.

cs.OS↗

arXiv2025

Compact dose delivery of laser-accelerated high-energy electron beams towards radiotherapy applications

The use of very high energy electron (VHEE) beams for radiotherapy has been actively studied for over two decades due to their advantageous dose distribution, deep penetration depth and great potential of ultra-high dose-rate irradiation. Recently, laser-plasma wakefield accelerator (LWFA) has emerged as a promising method for the compact generation of VHEE beams, due to its substantially higher accelerating gradients compared to traditional radio-frequency accelerators. However, how to compactly deliver the LWFA-based VHEE beams of relatively large energy spread and create a maximum dose deeply inside the body remains very challenging. In this article, we present a simple dose delivery scheme utilizing only two dipole magnets for LWFA-based VHEE treatment. By adjusting the magnet strengths, the electron beams can be guided along different angular trajectories towards a precise position as deep as 20 cm within a water phantom, creating a maximum dose over the target region and significantly reducing the entrance dose. Supported by Monte Carlo simulations, such a beam delivery approach is demonstrated to be insensitive to the beam energy spread and meanwhile capable of controlling precisely the dose-peak position in both lateral and longitudinal directions. As such, a uniform dose peak can be generated by the weighted sum of VHEE beams that reach different dose-peak depths. These results demonstrate that LWFA-based VHEE beams can be compactly delivered into a deep-seated tumor region in a controllable manner, thus advancing the development of the VHEE radiotherapy towards the practical clinical applications in the near future.

physics.med-ph↗

arXiv2024

Simulating continuum-based redshift measurement in the \textit{Roman's} High Latitude Spectroscopy Survey

We investigate the capability of the \textit{Nancy Grace Roman Space Telescope's (Roman)} Wide-Field Instrument (WFI) G150 slitless grism to detect red, quiescent galaxies based on the current reference survey. We simulate dispersed images for \textit{Roman} reference High-Latitude Spectroscopic Survey (HLSS) and analyze two-dimensional spectroscopic data using the grism Redshift and Line Analysis (\verb|Grizli|) software. This study focus on assessing \textit{Roman} grism's capability for continuum-level redshift measurement for a redshift range of $0.5 \leq z \leq 2.5$. The redshift recovery is assessed by setting three requirements of: $σ_z = \frac{\left|z-z_{\mathrm{true}}\right|}{1+z}\leq0.01$, signal-to-noise ratio (S/N) $\geq 5$ and the presence of a single dominant peak in redshift likelihood function. We find that, for quiescent galxaies, the reference HLSS can reach a redshift recovery completeness of $\geq50\%$ for F158 magnitude brighter than 20.2 mag. We also explore how different survey parameters, such as exposure time and the number of exposures, influence the accuracy and completeness of redshift recovery, providing insights that could optimize future survey strategies and enhance the scientific yield of the \textit{Roman} in cosmological research.

astro-ph.GA↗

arXiv2024

Zenix: Efficient Execution of Bulky Serverless Applications

Serverless computing, commonly offered as Function-as-a-Service, was initially designed for small, lean applications. However, there has been an increasing desire to run larger, more complex applications (what we call bulky applications) in a serverless manner. Existing strategies for enabling such applications are to either increase function sizes or to rewrite applications as DAGs of functions. These approaches cause significant resource wastage, manual efforts, and/or performance overhead. We argue that the root cause of these issues is today's function-centric serverless model, where a function is the resource allocation and scaling unit. We propose a new, resource-centric serverless-computing model for executing bulky applications in a resource- and performance-efficient way, and we build the Zenix serverless platform following this model. Our results show that Zenix reduces resource consumption by up to 90% compared to today's function-centric serverless systems, while improving performance by up to 64%.

cs.DC↗

arXiv2023

Experimental demonstration of mice tumor control with a laser-accelerated high-energy electron radiotherapy prototype

Radiotherapy using very-high-energy electron (VHEE) beams (50-300 MeV) has attracted considerable attention due to its advantageous dose deposition characteristics, enabling deep penetration and the potential for ultra-high dose rate treatment. One promising approach to compactly delivering these high energy electron beams in a cost-effective manner is laser wakefield acceleration (LWFA), which offers ultra-strong accelerating gradients. However, the transition from this concept to a functional machine intended for tumor treatment is still being investigated. Here we present the first self-developed prototype for LWFA-based VHEE radiotherapy, exhibiting high compactness (occupying less than 5 square meters) and high operational stability (validated over a period of one month). Subsequently, we employed this device to irradiate a tumor implanted in a mouse model. Following a dose delivery of $5.8\pm0.2$ Gy with precise tumor conformity, all irradiated mice exhibited pronounced control of tumor growth. For comparison, this tumor-control efficacy was similar to that achieved using commercial X-ray radiotherapy equipment operating at equivalent doses. These results demonstrate the potential of a compact laser-driven VHEE system for preclinical studies involving small animal models and its promising prospects for future clinical translation in cancer therapy.

physics.med-ph↗

arXiv2023

Galaxy bias in the era of LSST: perturbative bias expansions

Upcoming imaging surveys will allow for high signal-to-noise measurements of galaxy clustering at small scales. In this work, we present the results of the LSST bias challenge, the goal of which is to compare the performance of different nonlinear galaxy bias models in the context of LSST Y10 data. Specifically, we compare two perturbative approaches, Lagrangian perturbation theory (LPT) and Eulerian PT (EPT) to two variants of Hybrid Effective Field Theory (HEFT), with our fiducial implementation of these models including terms up to second order in the bias expansion as well as nonlocal bias and deviations from Poissonian stochasticity. We consider different simulated galaxy samples and test the performance of the bias models in a tomographic joint analysis of LSST-Y10-like galaxy clustering, galaxy-galaxy-lensing and cosmic shear. We find both HEFT methods as well as LPT and EPT combined with non-perturbative predictions for the matter power spectrum to yield unbiased constraints on cosmological parameters up to at least a maximal scale of $k_{\mathrm{max}}=0.4 \; \mathrm{Mpc}^{-1}$ for all samples considered, even in the presence of assembly bias. While we find that we can reduce the complexity of the bias model for HEFT without compromising fit accuracy, this is not generally the case for the perturbative models. We find significant detections of non-Poissonian stochasticity in all cases considered, and our analysis shows evidence that small-scale galaxy clustering predominantly improves constraints on galaxy bias rather than cosmological parameters. These results therefore suggest that the systematic uncertainties associated with current nonlinear bias models are likely to be subdominant compared to other sources of error for tomographic analyses of upcoming photometric surveys, which bodes well for future galaxy clustering analyses using these high signal-to-noise data. [abridged]

astro-ph.CO↗

arXiv2023

MSTFormer: Motion Inspired Spatial-temporal Transformer with Dynamic-aware Attention for long-term Vessel Trajectory Prediction

Incorporating the dynamics knowledge into the model is critical for achieving accurate trajectory prediction while considering the spatial and temporal characteristics of the vessel. However, existing methods rarely consider the underlying dynamics knowledge and directly use machine learning algorithms to predict the trajectories. Intuitively, the vessel's motions are following the laws of dynamics, e.g., the speed of a vessel decreases when turning a corner. Yet, it is challenging to combine dynamic knowledge and neural networks due to their inherent heterogeneity. Against this background, we propose MSTFormer, a motion inspired vessel trajectory prediction method based on Transformer. The contribution of this work is threefold. First, we design a data augmentation method to describe the spatial features and motion features of the trajectory. Second, we propose a Multi-headed Dynamic-aware Self-attention mechanism to focus on trajectory points with frequent motion transformations. Finally, we construct a knowledge-inspired loss function to further boost the performance of the model. Experimental results on real-world datasets show that our strategy not only effectively improves long-term predictive capability but also outperforms backbones on cornering data.The ablation analysis further confirms the efficacy of the proposed method. To the best of our knowledge, MSTFormer is the first neural network model for trajectory prediction fused with vessel motion dynamics, providing a worthwhile direction for future research.The source code is available at https://github.com/simple316/MSTFormer.

cs.LG↗

arXiv2022

Fringing Analysis and Simulation for the Vera C. Rubin Observatory's Legacy Survey of Space and Time

The presence of fringing in astronomical CCD images will impact photometric quality and measurements. Yet its impact on the Vera C. Rubin Observatory's Legacy Survey of Space and Time (LSST) has not been fully studied. We present a detailed study on fringing for Charge-Coupled Devices (CCDs) already implemented on the Rubin Observatory LSST Camera's focal plane. After making physical measurements and knowing the compositions, we have developed a model for the e2v CCDs. We present a method to fit for the internal height variation of the epoxy layer within the sensors based on fringing measurements in a laboratory setting. This method is generic enough that it can be easily modified to work for other CCDs. Using the derived fringing model, we successfully reproduce comparable fringing amplitudes that match the observed levels in images taken by existing telescopes with different optical designs. This model is then used to forecast the expected level of fringing in a single LSST y-band sky background exposure with Rubin telescope optics in the presence of a realistic time varying sky spectrum. The predicted fringing amplitude in LSST images ranges from $0.04\%$ to $0.2\%$ depending on the location of a CCD on the focal plane. We find that the predicted variation in surface brightness caused by fringing in LSST y-band skybackground images is about $0.6\ μ\rm{Jy}\ \rm{arcsec}^{-2}$, which is 40 times larger than the current measurement error. We conclude that it is necessary to include fringing correction in the Rubin's LSST image processing pipeline.

astro-ph.IM↗

arXiv2022

Clio: A Hardware-Software Co-Designed Disaggregated Memory System

Memory disaggregation has attracted great attention recently because of its benefits in efficient memory utilization and ease of management. So far, memory disaggregation research has all taken one of two approaches: building/emulating memory nodes using regular servers or building them using raw memory devices with no processing power. The former incurs higher monetary cost and faces tail latency and scalability limitations, while the latter introduces performance, security, and management problems. Server-based memory nodes and memory nodes with no processing power are two extreme approaches. We seek a sweet spot in the middle by proposing a hardware-based memory disaggregation solution that has the right amount of processing power at memory nodes. Furthermore, we take a clean-slate approach by starting from the requirements of memory disaggregation and designing a memory-disaggregation-native system. We built Clio, a disaggregated memory system that virtualizes, protects, and manages disaggregated memory at hardware-based memory nodes. The Clio hardware includes a new virtual memory system, a customized network system, and a framework for computation offloading. In building Clio, we not only co-design OS functionalities, hardware architecture, and the network system, but also co-design compute nodes and memory nodes. Our FPGA prototype of Clio demonstrates that each memory node can achieve 100 Gbps throughput and an end-to-end latency of 2.5 us at median and 3.2us at the 99th percentile. Clio also scales much better and has orders of magnitude lower tail latency than RDMA. It has 1.1x to 3.4x energy saving compared to CPU-based and SmartNIC-based disaggregated memory systems and is 2.7x faster than software-based SmartNIC solutions.

cs.DC↗