arXiv subjects

Shaoqing Zhang

Publications and source records attributed to Shaoqing Zhang.

17 recordsLinked to original sources

arXiv2026

Python-Fortran Hybrid Programming to Fuse AI and Physical Models: Examples of AI-LDA in climate and weather models (Hf2pMDA_v1.0)

AI provides an unprecedented opportunity for advancing physics numerical modeling including data assimilation, which is a highly efficient and critically-important tool for advancing our understanding on Earth system and its applications. At the same time, deep incorporation of AI and physical modeling can make great driving to advance AI by injecting it rich physics from long time physics-based modeling development. However, since such physics models are conventionally coded in Fortran and AI algorithms usually are conveniently designed in Python, difficulties exist to directly incorporate AI algorithms into physics models, vice versa. Here, based on the F2PY protocol, we have developed a procedure that implements an infrastructure which conveniently conducts Hf2pMDA to form a program entity so that AI algorithms and physical models can invoke mutually. As examples, within Hf2pMDA, a climate coupled data assimilation (CDA) system is naturally upgraded to a strongly CDA (SCDA) system, and a 1 km high-resolution weather DA system is conveniently implemented within a multi-layer downscaling model that has multiscale DA in different nesting layers. In the climate SCDA system, a coupled general circulation model (CGCM) and a multiscale filtering algorithm is integrated by a Python main controller (PMC) that calls Fortran CGCM components and Weakly-CDA modules as well as a data-trained SCDA algorithm by latent space autoencoder in Python. In the high-resolution weather DA system, the downscaled model consisting of traditional Fortran DA modules in all mother domains and Python AE DA algorithm in the central child domain is integrated by a PMC that organizes these components. With convenient realization of deep incorporation of any AI algorithm and physics model, the Hf2pMDA has a great potential to make progress on both AI and scientific modeling.

physics.ao-ph↗

arXiv2026

MemFine: Memory-Aware Fine-Grained Scheduling for MoE Training

The training of large-scale Mixture of Experts (MoE) models faces a critical memory bottleneck due to severe load imbalance caused by dynamic token routing. This imbalance leads to memory overflow on GPUs with limited capacity, constraining model scalability. Existing load balancing methods, which cap expert capacity, compromise model accuracy and fail on memory-constrained hardware. To address this, we propose MemFine, a memory-aware fine-grained scheduling framework for MoE training. MemFine decomposes the token distribution and expert computation into manageable chunks and employs a chunked recomputation strategy, dynamically optimized through a theoretical memory model to balance memory efficiency and throughput. Experiments demonstrate that MemFine reduces activation memory by 48.03% and improves throughput by 4.42% compared to full recomputation-based baselines, enabling stable large-scale MoE training on memory-limited GPUs.

cs.DC↗

arXiv2025

MoFa: A Unified Performance Modeling Framework for LLM Pretraining

The exponential growth in LLM scales, with parameters soaring from billions to trillions, has necessitated distributed pretraining across large clusters comprising thousands to tens of thousands of devices. While hybrid parallelization strategies enable such pretraining, the vast combinatorial strategy space introduces significant optimization challenges. Traditional manual tuning methods incur prohibitive trial-and-error costs, and existing performance modeling approaches exhibit critical limitations: they fail to comprehensively account for prevalent optimization features and ignore the substantial overhead imposed by essential fault tolerance mechanisms like checkpoint recovery in long-duration pretraining. To address these gaps, we propose MoFa, a novel pretraining performance modeling framework that unifies multi-dimensional optimization features and fault tolerance. MoFa incorporates an enhanced cost model to accurately capture the effects of key optimizations and integrates a fault tolerance model based on historical cluster reliability data. Besides, a MoFa-based tuning system is developed to explore optimal pretraining performance and potential bottlenecks in various scenarios. Extensive modeling evaluations demonstrate that MoFa can achieve high prediction accuracy across various scenarios. In addition, through comprehensive tuning experiments, our framework systematically reveals the key factors influencing pretraining performance under different configurations, which provides solid a priori guidance for LLM pretraining system design and deployment.

cs.DC↗

arXiv2025

Cracking the Code of Arctic Sea Ice: Why Models Fail to Predict Its Retreat?

Arctic sea ice is rapidly retreating due to global warming, and emerging evidence suggests that the rate of decline may have been underestimated. A key factor contributing to this underestimation is the coarse resolution of current climate models, which fail to accurately represent eddy floe interactions, climate extremes, and other critical small scale processes. Here, we elucidate the roles of these dynamics in accelerating sea ice melt and emphasize the need for higher resolution models to improve projections of Arctic sea ice.

physics.ao-ph↗

arXiv2025

Typhoon Tracks Regulated by Feedbacks of Fine-Scale Clouds to Environment

Accurate tropical cyclone (TC) track prediction is crucial for mitigating the catastrophic impacts of TCs on human life and the environment. Despite decades of research on tropical cyclone (TC) track prediction, large errors known as track forecast busts (TFBs) occur frequently, and their causes remain poorly understood. Here, we examine a few dozens of TCs using a unique TC downscaling strategy that can quantitatively assess the sensitivity of TC track on the strength of feedbacks of fine-scale clouds to environment. We show that as TFBs have a weaker environmental steering that favors scattering cumulonimbus clouds, capturing asymmetric distribution of planetary vorticity advection induced by such fine-scale clouds corrects TFBs by 60 percent. Our clear identification of such important TC track predictability source promises continuous improvement of TC track prediction as finer-scale TC clouds and their interactions with environment are better resolved as model larger-scale behaviors have improved.

physics.ao-ph↗

arXiv2025

XBOUND: Exploring Capability Boundaries of Device-Control Agents at the State Level

Recent advancements in vision-language models have increased interest in Device-Control Agents (DC agents) for managing graphical user interfaces (GUIs). With the growing complexity and integration of such agents into various applications, effective evaluation methods have become crucial. The current evaluation method for DC agents primarily focuses on the instruction level, providing the current state (e.g., screenshots) and past execution history to determine actions for target instructions, helping identify potential execution failures. However, in GUI environments, a single state may contain multiple interactive widgets, each linked to different instructions, presenting an opportunity for diverse actions based on various instruction targets. Evaluating the agent's performance solely at the instruction level may overlook the broader context of these interactions. To capture a more comprehensive view of agent performance, we propose a new evaluation method, XBOUND, to evaluate the accuracy of instruction completion on a per-state basis. XBOUND provides a state-level evaluation framework, serving as a tool to assess agents' capabilities within environmental states. Our evaluation yields several key insights: UI-TARS stands out as the strongest 7B model, current agents display a bimodal performance pattern in instruction unification, and sub-7B models remain limited in state mastery. We further identify GPT-based planning as a critical bottleneck, and show that grounding data mainly benefits action matching, while trajectory data is more effective for instruction unification.

cs.AI↗

arXiv2025

Interpretable Cross-Sphere Multiscale Deep Learning Predicts ENSO Skilfully Beyond 2 Years

El Niño-Southern Oscillation (ENSO) exerts global climate and societal impacts, but real-time prediction with lead times beyond one year remains challenging. Dynamical models suffer from large biases and uncertainties, while deep learning struggles with interpretability and multi-scale dynamics. Here, we introduce PTSTnet, an interpretable model that unifies dynamical processes and cross-scale spatiotemporal learning in an innovative neural-network framework with physics-encoding learning. PTSTnet produces interpretable predictions significantly outperforming state-of-the-art benchmarks with lead times beyond 24 months, providing physical insights into error propagation in ocean-atmosphere interactions. PTSTnet learns feature representations with physical consistency from sparse data to tackle inherent multi-scale and multi-physics challenges underlying ocean-atmosphere processes, thereby inherently enhancing long-term prediction skill. Our successful realizations mark substantial steps forward in interpretable insights into innovative neural ocean modelling.

physics.ao-ph↗

arXiv2025

Ekman Theory with Damping

The observed Ekman spirals in the ocean are always "flatter" than that predicted by the classic theory. We propose that the universal flattening of Ekman spiral is mainly due to the damping associated with turbulent dissipation. Analytical solutions and numerical simulations show convincingly a better fitting between the new theory and observations. Most importantly, the new theory indicates that the damping can lead to weakened Ekman transport and pumping, with the latter not only driven by the curl but also the divergence of wind stress. Under a modest damping, the Ekman transport along 26.5°N will be ~0.4 Sv (12%) smaller than that predicted by the classic theory. Hence, the damping due to turbulent dissipation can noticeably affect the wind-driven circulation in the upper ocean.

physics.ao-ph↗

arXiv2025

Look Before You Leap: Enhancing Attention and Vigilance Regarding Harmful Content with GuidelineLLM

Despite being empowered with alignment mechanisms, large language models (LLMs) are increasingly vulnerable to emerging jailbreak attacks that can compromise their alignment mechanisms. This vulnerability poses significant risks to real-world applications. Existing work faces challenges in both training efficiency and generalization capabilities (i.e., Reinforcement Learning from Human Feedback and Red-Teaming). Developing effective strategies to enable LLMs to resist continuously evolving jailbreak attempts represents a significant challenge. To address this challenge, we propose a novel defensive paradigm called GuidelineLLM, which assists LLMs in recognizing queries that may have harmful content. Before LLMs respond to a query, GuidelineLLM first identifies potential risks associated with the query, summarizes these risks into guideline suggestions, and then feeds these guidelines to the responding LLMs. Importantly, our approach eliminates the necessity for additional safety fine-tuning of the LLMs themselves; only the GuidelineLLM requires fine-tuning. This characteristic enhances the general applicability of GuidelineLLM across various LLMs. Experimental results demonstrate that GuidelineLLM can significantly reduce the attack success rate (ASR) against LLM (an average reduction of 34.17\% ASR) while maintaining the usefulness of LLM in handling benign queries. The code is available at https://github.com/sqzhang-lazy/GuidelineLLM.

cs.CL↗

arXiv2025

Organization of Historical Oceanic Overturnings on Cross-Sphere Climate Signals

The global ocean meridional overturning circulation (GMOC) is central for ocean transport and climate variations. However, a comprehensive picture of its historical mean state and variability remains vague due to limitations in modelling and observing systems. Incorporating observations into models offers a viable approach to reconstructing climate history, yet achieving coherent estimates of GMOC has proven challenging due to difficulties in harmonizing ocean stratification. Here, we demonstrate that applying multiscale data assimilation scheme that integrates atmospheric and oceanic observations into multiple coupled models in a dynamically consistent way, the global ocean currents and GMOC over the past 80 years are retrieved. While the major historic events are printed in variability of the rebuilt GMOC, the timeseries of multisphere 3-dimensional physical variables representing the realistic historical evolution enable us to advance understanding of mechanisms of climate signal propagation cross spheres and give birth to Artificial Intelligence coupled big models, thus advancing the Earth science.

physics.ao-ph↗

arXiv2025

AI Models Still Lag Behind Traditional Numerical Models in Predicting Sudden-Turning Typhoons

Given the interpretability, accuracy, and stability of numerical weather prediction (NWP) models, current operational weather forecasting relies heavily on the NWP approach. In the past two years, the rapid development of Artificial Intelligence (AI) has provided an alternative solution for medium-range (1-10 days) weather forecasting. Bi et al. (2023) (hereafter Bi23) introduced the first AI-based weather prediction (AIWP) model in China, named Pangu-Weather, which offers fast prediction without compromising accuracy. In their work, Bi23 made notable claims regarding its effectiveness in extreme weather predictions. However, this claim lacks persuasiveness because the extreme nature of the two tropical cyclones (TCs) examples presented in Bi23, namely Typhoon Kong-rey and Typhoon Yutu, stems primarily from their intensities rather than their moving paths. Their claim may mislead into another meaning which is that Pangu-Weather works well in predicting unusual typhoon paths, which was not explicitly analyzed. Here, we reassess Pangu-Weather's ability to predict extreme TC trajectories from 2020-2024. Results reveal that while Pangu-Weather overall outperforms NWP models in predicting tropical cyclone (TC) tracks, it falls short in accurately predicting the rarely observed sudden-turning tracks, such as Typhoon Khanun in 2023. We argue that current AIWP models still lag behind traditional NWP models in predicting such rare extreme events in medium-range forecasts.

physics.ao-ph↗

arXiv2024

Dynamic Planning for LLM-based Graphical User Interface Automation

The advent of large language models (LLMs) has spurred considerable interest in advancing autonomous LLMs-based agents, particularly in intriguing applications within smartphone graphical user interfaces (GUIs). When presented with a task goal, these agents typically emulate human actions within a GUI environment until the task is completed. However, a key challenge lies in devising effective plans to guide action prediction in GUI tasks, though planning have been widely recognized as effective for decomposing complex tasks into a series of steps. Specifically, given the dynamic nature of environmental GUIs following action execution, it is crucial to dynamically adapt plans based on environmental feedback and action history.We show that the widely-used ReAct approach fails due to the excessively long historical dialogues. To address this challenge, we propose a novel approach called Dynamic Planning of Thoughts (D-PoT) for LLM-based GUI agents.D-PoT involves the dynamic adjustment of planning based on the environmental feedback and execution history. Experimental results reveal that the proposed D-PoT significantly surpassed the strong GPT-4V baseline by +12.7% (34.66% $\rightarrow$ 47.36%) in accuracy. The analysis highlights the generality of dynamic planning in different backbone LLMs, as well as the benefits in mitigating hallucinations and adapting to unseen tasks. Code is available at https://github.com/sqzhang-lazy/D-PoT.

cs.AI↗

arXiv2024

Multi-Programming Language Sandbox for LLMs

We introduce MPLSandbox, an out-of-the-box multi-programming language sandbox designed to provide unified and comprehensive feedback from compiler and analysis tools for Large Language Models (LLMs). It can automatically identify the programming language of the code, compiling and executing it within an isolated sub-sandbox to ensure safety and stability. In addition, MPLSandbox also integrates both traditional and LLM-based code analysis tools, providing a comprehensive analysis of generated code. MPLSandbox can be effortlessly integrated into the training and deployment of LLMs to improve the quality and correctness of their generated code. It also helps researchers streamline their workflows for various LLM-based code-related tasks, reducing the development cost. To validate the effectiveness of MPLSandbox, we integrate it into training and deployment approaches, and also employ it to optimize workflows for a wide range of real-world code-related tasks. Our goal is to enhance researcher productivity on LLM-based code-related tasks by simplifying and automating workflows through delegation to MPLSandbox.

cs.SE↗

arXiv2024

Kilometer-Level Coupled Modeling Using 40 Million Cores: An Eight-Year Journey of Model Development

With current and future leading systems adopting heterogeneous architectures, adapting existing models for heterogeneous supercomputers is of urgent need for improving model resolution and reducing modeling uncertainty. This paper presents our three-week effort on porting a complex earth system model, CESM 2.2, to a 40-million-core Sunway supercomputer. Taking a non-intrusive approach that tries to minimizes manual code modifications, our project tries to achieve both improvement of performance and consistency of the model code. By using a hierarchical grid system and an OpenMP-based offloading toolkit, our porting and parallelization effort covers over 80% of the code, and achieves a simulation speed of 340 SDPD (simulated days per day) for 5-km atmosphere, 265 SDPD for 3-km ocean, and 222 SDPD for a coupled model, thus making multi-year or even multi-decadal experiments at such high resolution possible.

cs.DC↗

arXiv2023

Dual-Propagation-Feature Fusion Enhanced Neural CSI Compression for Massive MIMO

Due to the ability of feature extraction, deep learning (DL)-based methods have been recently applied to channel state information (CSI) compression feedback in massive multiple-input multiple-output (MIMO) systems. Existing DL-based CSI compression methods are usually effective in extracting a certain type of features in the CSI. However, the CSI usually contains two types of propagation features, i.g., non-line-of-sight (NLOS) propagation-path feature and dominant propagation-path feature, especially in channel environments with rich scatterers. To fully extract the both propagation features and learn a dual-feature representation for CSI, this paper proposes a dual-feature-fusion neural network (NN), referred to as DuffinNet. The proposed DuffinNet adopts a parallel structure with a convolutional neural network (CNN) and an attention-empowered neural network (ANN) to respectively extract different features in the CSI, and then explores their interplay by a fusion NN. Built upon this proposed DuffinNet, a new encoder-decoder framework is developed, referred to as Duffin-CsiNet, for improving the end-to-end performance of CSI compression and reconstruction. To facilitate the application of Duffin-CsiNet in practice, this paper also presents a two-stage approach for codeword quantization of the CSI feedback. Besides, a transfer learning-based strategy is introduced to improve the generalization of Duffin-CsiNet, which enables the network to be applied to new propagation environments. Simulation results illustrate that the proposed Duffin-CsiNet noticeably outperforms the existing DL-based methods in terms of reconstruction performance, encoder complexity, and network convergence, validating the effectiveness of the proposed dual-feature fusion design.

cs.IT↗

arXiv2022

Deep CSI Compression for Massive MIMO: A Self-information Model-driven Neural Network

In order to fully exploit the advantages of massive multiple-input multiple-output (mMIMO), it is critical for the transmitter to accurately acquire the channel state information (CSI). Deep learning (DL)-based methods have been proposed for CSI compression and feedback to the transmitter. Although most existing DL-based methods consider the CSI matrix as an image, structural features of the CSI image are rarely exploited in neural network design. As such, we propose a model of self-information that dynamically measures the amount of information contained in each patch of a CSI image from the perspective of structural features. Then, by applying the self-information model, we propose a model-and-data-driven network for CSI compression and feedback, namely IdasNet. The IdasNet includes the design of a module of self-information deletion and selection (IDAS), an encoder of informative feature compression (IFC), and a decoder of informative feature recovery (IFR). In particular, the model-driven module of IDAS pre-compresses the CSI image by removing informative redundancy in terms of the self-information. The encoder of IFC then conducts feature compression to the pre-compressed CSI image and generates a feature codeword which contains two components, i.e., codeword values and position indices of the codeword values. Subsequently, the IFR decoder decouples the codeword values as well as position indices to recover the CSI image. Experimental results verify that the proposed IdasNet noticeably outperforms existing DL-based networks under various compression ratios while it has the number of network parameters reduced by orders-of-magnitude compared with various existing methods.

cs.IT↗

arXiv2022

Data Augmentation Empowered Neural Precoding for Multiuser MIMO with MMSE Model

Precoding design exploiting deep learning methods has been widely studied for multiuser multiple-input multiple-output (MU-MIMO) systems. However, conventional neural precoding design applies black-box-based neural networks which are less interpretable. In this paper, we propose a deep learning-based precoding method based on an interpretable design of a neural precoding network, namely iPNet. In particular, the iPNet mimics the classic minimum mean-squared error (MMSE) precoding and approximates the matrix inversion in the design of the neural network architecture. Specifically, the proposed iPNet consists of a model-driven component network, responsible for augmenting the input channel state information (CSI), and a data-driven sub-network, responsible for precoding calculation from this augmented CSI. The latter data-driven module is explicitly interpreted as an unsupervised learner of the MMSE precoder. Simulation results show that by exploiting the augmented CSI, the proposed iPNet achieves noticeable performance gain over existing black-box designs and also exhibits enhanced generalizability against CSI mismatches.

cs.IT↗