arXiv subjects

Zhi Wang

Publications and source records attributed to Zhi Wang.

At least 127 records · Page 7Linked to original sources

arXiv2025

MoETTA: Test-Time Adaptation Under Mixed Distribution Shifts with MoE-LayerNorm

Test-Time adaptation (TTA) has proven effective in mitigating performance drops under single-domain distribution shifts by updating model parameters during inference. However, real-world deployments often involve mixed distribution shifts, where test samples are affected by diverse and potentially conflicting domain factors, posing significant challenges even for SOTA TTA methods. A key limitation in existing approaches is their reliance on a unified adaptation path, which fails to account for the fact that optimal gradient directions can vary significantly across different domains. Moreover, current benchmarks focus only on synthetic or homogeneous shifts, failing to capture the complexity of real-world heterogeneous mixed distribution shifts. To address this, we propose MoETTA, a novel entropy-based TTA framework that integrates the Mixture-of-Experts (MoE) architecture. Rather than enforcing a single parameter update rule for all test samples, MoETTA introduces a set of structurally decoupled experts, enabling adaptation along diverse gradient directions. This design allows the model to better accommodate heterogeneous shifts through flexible and disentangled parameter updates. To simulate realistic deployment conditions, we introduce two new benchmarks: potpourri and potpourri+. While classical settings focus solely on synthetic corruptions, potpourri encompasses a broader range of domain shifts--including natural, artistic, and adversarial distortions--capturing more realistic deployment challenges. Additionally, potpourri+ further includes source-domain samples to evaluate robustness against catastrophic forgetting. Extensive experiments across three mixed distribution shifts settings show that MoETTA consistently outperforms strong baselines, establishing SOTA performance and highlighting the benefit of modeling multiple adaptation directions via expert-level diversity.

cs.LG↗

arXiv2025

FaSDiff: Balancing Perception and Semantics in Face Compression via Stable Diffusion Priors

With the increasing deployment of facial image data across a wide range of applications, efficient compression tailored to facial semantics has become critical for both storage and transmission. While recent learning-based face image compression methods have achieved promising results, they often suffer from degraded reconstruction quality at low bit rates. Directly applying diffusion-based generative priors to this task leads to suboptimal performance in downstream machine vision tasks, primarily due to poor preservation of high-frequency details. In this work, we propose FaSDiff (\textbf{Fa}cial Image Compression with a \textbf{S}table \textbf{Diff}usion Prior), a novel diffusion-driven compression framework designed to enhance both visual fidelity and semantic consistency. FaSDiff incorporates a high-frequency-sensitive compressor to capture fine-grained details and generate robust visual prompts for guiding the diffusion model. To address low-frequency degradation, we further introduce a hybrid low-frequency enhancement module that disentangles and preserves semantic structures, enabling stable modulation of the diffusion prior during reconstruction. By jointly optimizing perceptual quality and semantic preservation, FaSDiff effectively balances human visual fidelity and machine vision accuracy. Extensive experiments demonstrate that FaSDiff outperforms state-of-the-art methods in both perceptual metrics and downstream task performance.

cs.CV↗

arXiv2025

Microscopic origin of the entropy of de Sitter spacetime

We construct an infinite family of semiclassical de Sitter (dS) microstates, realized as backreacted geometries of dS spacetime with a constant tension thin-shell brane located outside the dS event horizon. We further show that wormhole contributions to the semiclassical Euclidean gravitational path integral lead to universal nonperturbative overlaps between these microstates. By evaluating the nonperturbative overlaps, we count the dimension of the Hilbert space spanned by the semiclassical dS microstates and find that it precisely equals the exponential of the Gibbons-Hawking entropy of dS spacetime. Our construction thus provides a state-counting derivation for Gibbons-Hawking entropy of dS spacetime.

hep-th↗

arXiv2025

Bridging Lifelong and Multi-Task Representation Learning via Algorithm and Complexity Measure

In lifelong learning, a learner faces a sequence of tasks with shared structure and aims to identify and leverage it to accelerate learning. We study the setting where such structure is captured by a common representation of data. Unlike multi-task learning or learning-to-learn, where tasks are available upfront to learn the representation, lifelong learning requires the learner to make use of its existing knowledge while continually gathering partial information in an online fashion. In this paper, we consider a generalized framework of lifelong representation learning. We propose a simple algorithm that uses multi-task empirical risk minimization as a subroutine and establish a sample complexity bound based on a new notion we introduce--the task-eluder dimension. Our result applies to a wide range of learning problems involving general function classes. As concrete examples, we instantiate our result on classification and regression tasks under noise.

cs.LG↗

arXiv2025

Quantum gravity corrections to the spontaneous excitation of an accelerated atom interacting with a quantum scalar field

The Generalized Uncertainty Principle (GUP) extends the Heisenberg Uncertainty Principle (HUP) by suggesting a minimum observable scale that includes the effects of quantum gravity, which is supposed to potentially result in observable effects far below the Planck energy scale, providing us the opportunity to explore the theory of quantum gravity through physical processes at low energy scale. In present work, we study the corrections induced by the GUP to the spontaneous radiation properties of a two-level atom interacting with a real massless scalar quantum field based on the DDC formalism. The GUP alters the correlation function of the scalar field, consequently affecting the radiative properties of atoms. We calculate the rate of change in the mean atomic energy for an atom undergoing inertial motion, uniform acceleration, and uniform circular motion. We show that the GUP can modify the spontaneous emission rate of an excited-state atom in inertial motion; however, it does not alter the stability of the ground-state atom in vacuum. For an atom in uniformly accelerated and uniformly circular motions, the GUP can change both its spontaneous emission and excitation rates; moreover, the corrections caused by the GUP contains the terms proportional to $βa^2$ or $βa^3$, suggesting that the proper acceleration $a$ of an atom in non-inertial motions could significantly amplify the effect of the GUP on the spontaneous transition rates of the atom.

hep-ph↗

arXiv2025

Investigation of the short-range correlation and its induced high-momentum tail by photon emission from intermediate energy heavy-ion reactions

The effect of high-momentum tail (HMT) in nucleon momentum distribution of neutron-rich nuclei has been investigated by employing proton-neutron bremsstrahlung photons in the reactions of $^{124}$Sn +$ ^{124}$Sn, $^{112}$Sn + $^{112}$Sn and $^{197}$Au + $^{197}$Au at intermediate energies based on the IBUU transport model. With the time evolution of photon multiplicity and the double differential probability of photon production, we show the emission of bremsstrahlung photons is sensitive to the high-momentum component in nucleon momentum distribution of neutron-rich nuclei, and the HMT can lead to an obvious increase of hard photon production from proton-neutron bremsstrahlung collisions. Finally, the ratio of double differential photon production probability is examined as possible probe of the HMT in nucleon momentum distribution of neutron-rich nuclei.

nucl-th↗

arXiv2025

Mixture-of-Experts Meets In-Context Reinforcement Learning

In-context reinforcement learning (ICRL) has emerged as a promising paradigm for adapting RL agents to downstream tasks through prompt conditioning. However, two notable challenges remain in fully harnessing in-context learning within RL domains: the intrinsic multi-modality of the state-action-reward data and the diverse, heterogeneous nature of decision tasks. To tackle these challenges, we propose T2MIR (Token- and Task-wise MoE for In-context RL), an innovative framework that introduces architectural advances of mixture-of-experts (MoE) into transformer-based decision models. T2MIR substitutes the feedforward layer with two parallel layers: a token-wise MoE that captures distinct semantics of input tokens across multiple modalities, and a task-wise MoE that routes diverse tasks to specialized experts for managing a broad task distribution with alleviated gradient conflicts. To enhance task-wise routing, we introduce a contrastive learning method that maximizes the mutual information between the task and its router representation, enabling more precise capture of task-relevant information. The outputs of two MoE components are concatenated and fed into the next layer. Comprehensive experiments show that T2MIR significantly facilitates in-context learning capacity and outperforms various types of baselines. We bring the potential and promise of MoE to ICRL, offering a simple and scalable architectural enhancement to advance ICRL one step closer toward achievements in language and vision communities. Our code is available at https://github.com/NJU-RL/T2MIR.

cs.LG↗

arXiv2025

A GPU-based Monte Carlo framework for IMRT QA using EPID transit dosimetry

Purpose: We presented a GPU-based MC framework, ARCHER-EPID, specifically designed for EPID transit dosimetry, with improving accuracy and efficiency. Methods: A comprehensive MC framework was developed to perform full radiation transport simulations through three distinct zones: a detailed linear accelerator head model, a CT-based patient/phantom geometry, and a realistic, multi-layered EPID model. To convert the simulated absorbed dose to a realistic detector signal, a dose-response correction model was implemented. The framework was validated by comparing simulations against experimental measurements for 25 IMRT fields delivered to both a solid water phantom and a anthropomorphic phantom. Agreement was quantified using Gamma analysis. Results: The GPU-accelerated ARCHER-EPID framework can complete the simulation for a complex IMRT field in about 90 seconds. A 2D correction factor lookup table is generated by parameterizing radiological thickness and effective field size to account for the EPID's energy-dependent response. The data revealed that for small fields, beam hardening is the dominant effect, while for large fields, the contribution from patient-generated scatter overwhelms this effect. The average 2D gamma passing rates (3%/3 mm criteria) between simulation and measurements are 98.43% for the solid water phantom and 97.86% for the anthropomorphic phantom, respectively. Visual comparison of the images and dose profiles between simulation and measurements show a high degree of agreement. Conclusions: We have successfully developed and validated a GPU-based MC framework that provides gold-standard accuracy for EPID transit dosimetry in radiotherapy. The results demonstrate that our proposed method has potential for routine application in PSQA.

physics.med-ph↗

arXiv2025

First experimental proof of PET imaging based on multi-anode MCP-PMTs with Cherenkov radiator-integrated window

Improving the coincidence time resolution (CTR) of time-of-flight positron emission tomography (TOF-PET) systems to achieve a higher signal-to-noise ratio (SNR) gain or even direct positron emission imaging (dPEI) is of paramount importance for many advanced new clinical applications of PET imaging. This places higher demands on the timing performance of all aspects of PET systems. One effective approach is to use microchannel plate photomultiplier tubes (MCP-PMTs) for prompt Cherenkov photon detection. In this study, we developed a dual-module Cherenkov PET imaging experimental platform, utilising our proprietary 8 * 8-anode Cherenkov radiator-integrated window MCP-PMTs in combination with custom-designed multi-channel electronics, and designed a specific calibration and correction method for the platform. Using this platform, a CTR of 103 ps FWHM was achieved. We overcame the limitations of single-anode detectors in previous experiments, significantly enhanced imaging efficiency and achieved module-level Cherenkov PET imaging for the first time. Imaging experiments involving radioactive sources and phantoms of various shapes and types were conducted, which preliminarily validated the feasibility and advancement of this imaging method. In addition, the effects of normalisation correction and the interaction probability between the gamma rays and the MCP on the images and experimental results were analysed and verified.

physics.ins-det↗

arXiv2025

Interpretable Machine Learning for Urban Heat Mitigation: Attribution and Weighting of Multi-Scale Drivers

Urban heat islands (UHIs) are often accentuated during heat waves (HWs) and pose a public health risk. Mitigating UHIs requires urban planners to first estimate how urban heat is influenced by different land use types (LUTs) and drivers across scales - from synoptic-scale climatic background processes to small-scale urban- and scale-bridging features. This study proposes to classify these drivers into driving (D), urban (U), and local (L) features, respectively. To increase interpretability and enhance computation efficiency, a LUT-distinguishing machine learning approach is proposed as a fast emulator for Weather Research and Forecasting model (WRF) coupled to the Noah land surface model (LSM) to predict ground- (TSK) and 2-meter air temperature (T2). Using random forest regression (RFR) with extreme gradient boosting (XGB) trained on WRF output over Zurich, Switzerland, during heatwave (HW) periods in 2017 and 2019, this study proposes LUT-based (LB) models that categorize features by scales and practical controllability, allowing optional categorical weighting. This approach enables category-specific feature ranking and sensitivity estimation of T2 and TSK to most important small-scale drivers - most notably surface emissivity, albedo, and leaf area index (LAI). Models employing the LB framework are statistically significantly more accurate than models that do not, with higher performance when more HW data is included in training. With RFR-XGB robustly performing optimal with unit weights, the method substantially increase interpretability. Despite the needs to reduce uncertainties and test the method on other cities, the proposed approach offers urban planners a direct framework for feasibility-centered UHI mitigation assessment.

physics.ao-ph↗

arXiv2025

A Unified Predictive and Generative Solution for Liquid Electrolyte Formulation

Liquid electrolytes are critical components of next-generation energy storage systems, enabling fast ion transport, minimizing interfacial resistance, and ensuring electrochemical stability for long-term battery performance. However, measuring electrolyte properties and designing formulations remain experimentally and computationally expensive. In this work, we present a unified framework for designing liquid electrolyte formulation, integrating a forward predictive model with an inverse generative approach. Leveraging both computational and experimental data collected from literature and extensive molecular simulations, we train a predictive model capable of accurately estimating electrolyte properties from ionic conductivity to solvation structure. Our physics-informed architecture preserves permutation invariance and incorporates empirical dependencies on temperature and salt concentration, making it broadly applicable to property prediction tasks across molecular mixtures. Furthermore, we introduce -- to the best of our knowledge -- the first generative machine learning framework for molecular mixture design, demonstrated on electrolyte systems. This framework supports multi-condition-constrained generation, addressing the inherently multi-objective nature of materials design. As a proof of concept, we experimentally identified three liquid electrolytes with both high ionic conductivity and anion-concentrated solvation structure. This unified framework advances data-driven electrolyte design and can be readily extended to other complex chemical systems beyond electrolytes.

cond-mat.mtrl-sci↗

arXiv2025

ColNet: Collaborative Optimization in Decentralized Federated Multi-task Learning Systems

The integration of Federated Learning (FL) and Multi-Task Learning (MTL) has been explored to address client heterogeneity, with Federated Multi-Task Learning (FMTL) treating each client as a distinct task. However, most existing research focuses on data heterogeneity (e.g., addressing non-IID data) rather than task heterogeneity, where clients solve fundamentally different tasks. Additionally, much of the work relies on centralized settings with a server managing the federation, leaving the more challenging domain of decentralized FMTL largely unexplored. Thus, this work bridges this gap by proposing ColNet, a framework designed for heterogeneous tasks in decentralized federated environments. ColNet partitions models into a backbone and task-specific heads, and uses adaptive clustering based on model and data sensitivity to form task-coherent client groups. Backbones are averaged within groups, and group leaders perform hyper-conflict-averse cross-group aggregation. Across datasets and federations, ColNet outperforms competing schemes under label and task heterogeneity and shows robustness to poisoning attacks.

cs.LG↗

arXiv2025

SP-VLA: A Joint Model Scheduling and Token Pruning Approach for VLA Model Acceleration

Vision-Language-Action (VLA) models have attracted increasing attention for their strong control capabilities. However, their high computational cost and low execution frequency hinder their suitability for real-time tasks such as robotic manipulation and autonomous navigation. Existing VLA acceleration methods primarily focus on structural optimization, overlooking the fact that these models operate in sequential decision-making environments. As a result, temporal redundancy in sequential action generation and spatial redundancy in visual input remain unaddressed. To this end, we propose SP-VLA, a unified framework that accelerates VLA models by jointly scheduling models and pruning tokens. Specifically, we design an action-aware model scheduling mechanism that reduces temporal redundancy by dynamically switching between VLA model and a lightweight generator. Inspired by the human motion pattern of focusing on key decision points while relying on intuition for other actions, we categorize VLA actions into deliberative and intuitive, assigning the former to the VLA model and the latter to the lightweight generator, enabling frequency-adaptive execution through collaborative model scheduling. To address spatial redundancy, we further develop a spatio-semantic dual-aware token pruning method. Tokens are classified into spatial and semantic types and pruned based on their dual-aware importance to accelerate VLA inference. These two mechanisms work jointly to guide the VLA in focusing on critical actions and salient visual information, achieving effective acceleration while maintaining high accuracy. Extensive experiments show that our method achieves 1.5$\times$ lossless acceleration in LIBERO and 2.4$\times$ in SimplerEnv, with up to 6% average performance gain. Inference frequency and latency improve by 2.2$\times$ in SimplerEnv and 1.4$\times$ in LIBERO.

cs.CV↗

arXiv2025

DescribeEarth: Describe Anything for Remote Sensing Images

Automated textual description of remote sensing images is crucial for unlocking their full potential in diverse applications, from environmental monitoring to urban planning and disaster management. However, existing studies in remote sensing image captioning primarily focus on the image level, lacking object-level fine-grained interpretation, which prevents the full utilization and transformation of the rich semantic and structural information contained in remote sensing images. To address this limitation, we propose Geo-DLC, a novel task of object-level fine-grained image captioning for remote sensing. To support this task, we construct DE-Dataset, a large-scale dataset contains 25 categories and 261,806 annotated instances with detailed descriptions of object attributes, relationships, and contexts. Furthermore, we introduce DE-Benchmark, a LLM-assisted question-answering based evaluation suite designed to systematically measure model capabilities on the Geo-DLC task. We also present DescribeEarth, a Multi-modal Large Language Model (MLLM) architecture explicitly designed for Geo-DLC, which integrates a scale-adaptive focal strategy and a domain-guided fusion module leveraging remote sensing vision-language model features to encode high-resolution details and remote sensing category priors while maintaining global context. Our DescribeEarth model consistently outperforms state-of-the-art general MLLMs on DE-Benchmark, demonstrating superior factual accuracy, descriptive richness, and grammatical soundness, particularly in capturing intrinsic object features and surrounding environmental attributes across simple, complex, and even out-of-distribution remote sensing scenarios. All data, code and weights are released at https://github.com/earth-insights/DescribeEarth.

cs.CV↗

arXiv2025

Beyond A Single AI Cluster: A Survey of Decentralized LLM Training

The emergence of large language models (LLMs) has revolutionized AI development, yet the resource demands beyond a single cluster or even datacenter, limiting accessibility to well-resourced organizations. Decentralized training has emerged as a promising paradigm to leverage dispersed resources across clusters, datacenters and regions, offering the potential to democratize LLM development for broader communities. As the first comprehensive exploration of this emerging field, we present decentralized LLM training as a resource-driven paradigm and categorize existing efforts into community-driven and organizational approaches. We further clarify this through: (1) a comparison with related paradigms, (2) a characterization of decentralized resources, and (3) a taxonomy of recent advancements. We also provide up-to-date case studies and outline future directions to advance research in decentralized LLM training.

cs.DC↗

arXiv2025

Conf-Profile: A Confidence-Driven Reasoning Paradigm for Label-Free User Profiling

User profiling, as a core technique for user understanding, aims to infer structural attributes from user information. Large Language Models (LLMs) provide a promising avenue for user profiling, yet the progress is hindered by the lack of comprehensive benchmarks. To bridge this gap, we propose ProfileBench, an industrial benchmark derived from a real-world video platform, encompassing heterogeneous user data and a well-structured profiling taxonomy. However, the profiling task remains challenging due to the difficulty of collecting large-scale ground-truth labels, and the heterogeneous and noisy user information can compromise the reliability of LLMs. To approach label-free and reliable user profiling, we propose a Confidence-driven Profile reasoning framework Conf-Profile, featuring a two-stage paradigm. We first synthesize high-quality labels by leveraging advanced LLMs with confidence hints, followed by confidence-weighted voting for accuracy improvement and confidence calibration for a balanced distribution. The multiple profile results, rationales, and confidence scores are aggregated and distilled into a lightweight LLM. We further enhance the reasoning ability via confidence-guided unsupervised reinforcement learning, which exploits confidence for difficulty filtering, quasi-ground truth voting, and reward weighting. Experimental results demonstrate that Conf-Profile delivers substantial performance through the two-stage training, improving F1 by 13.97 on Qwen3-8B.

cs.AI↗

arXiv2025

Quantization Meets OOD: Generalizable Quantization-aware Training from a Flatness Perspective

Current quantization-aware training (QAT) methods primarily focus on enhancing the performance of quantized models on in-distribution (I.D) data, while overlooking the potential performance degradation on out-of-distribution (OOD) data. In this paper, we first substantiate this problem through rigorous experiment, showing that QAT can lead to a significant OOD generalization performance degradation. Further, we find the contradiction between the perspective that flatness of loss landscape gives rise to superior OOD generalization and the phenomenon that QAT lead to a sharp loss landscape, can cause the above problem. Therefore, we propose a flatness-oriented QAT method, FQAT, to achieve generalizable QAT. Specifically, i) FQAT introduces a layer-wise freezing mechanism to mitigate the gradient conflict issue between dual optimization objectives (i.e., vanilla QAT and flatness). ii) FQAT proposes an disorder-guided adaptive freezing algorithm to dynamically determines which layers to freeze at each training step, effectively addressing the challenges caused by interference between layers. A gradient disorder metric is designed to help the algorithm identify unstable layers during training. Extensive experiments on influential OOD benchmark demonstrate the superiority of our method over state-of-the-art baselines under both I.D and OOD image classification tasks.

cs.CV↗

arXiv2025

Annotation-Free Open-Vocabulary Segmentation for Remote-Sensing Images

Semantic segmentation of remote sensing (RS) images is pivotal for comprehensive Earth observation, but the demand for interpreting new object categories, coupled with the high expense of manual annotation, poses significant challenges. Although open-vocabulary semantic segmentation (OVSS) offers a promising solution, existing frameworks designed for natural images are insufficient for the unique complexities of RS data. They struggle with vast scale variations and fine-grained details, and their adaptation often relies on extensive, costly annotations. To address this critical gap, this paper introduces SegEarth-OV, the first framework for annotation-free open-vocabulary segmentation of RS images. Specifically, we propose SimFeatUp, a universal upsampler that robustly restores high-resolution spatial details from coarse features, correcting distorted target shapes without any task-specific post-training. We also present a simple yet effective Global Bias Alleviation operation to subtract the inherent global context from patch features, significantly enhancing local semantic fidelity. These components empower SegEarth-OV to effectively harness the rich semantics of pre-trained VLMs, making OVSS possible in optical RS contexts. Furthermore, to extend the framework's universality to other challenging RS modalities like SAR images, where large-scale VLMs are unavailable and expensive to create, we introduce AlignEarth, which is a distillation-based strategy and can efficiently transfer semantic knowledge from an optical VLM encoder to an SAR encoder, bypassing the need to build SAR foundation models from scratch and enabling universal OVSS across diverse sensor types. Extensive experiments on both optical and SAR datasets validate that SegEarth-OV can achieve dramatic improvements over the SOTA methods, establishing a robust foundation for annotation-free and open-world Earth observation.

cs.CV↗