arXiv subjects

Ming Zhao

Publications and source records attributed to Ming Zhao.

At least 37 records · Page 2Linked to original sources

arXiv2025

Polarization Multiplexed Metalens Array Optical Chip for High-Performance LWIR Polarimetric Camera

Compared with traditional infrared thermal imaging, polarimetric imaging provides additional polarization information, which effectively enhances object contours and image contrast, with broad application in both military and civilian domains. However, the traditional long-wave infrared polarimetric camera suffers from severe thermal noise, low sensitivity and limited detection accuracy. To address the aforementioned problems, a novel cooled LWIR polarimetric camera based on an achromatic polarization multiplexed germanium-based metalens array optical chip is reported in this paper, enabling high-precision division of focal plane linearly polarimetric imaging. The proposed system demonstrates high-precision linearly polarimetric imaging, with the metalens array achieving an average transmittance of 84.7% across the 8.4~11.6μm band and a polarization extinction ratio exceeding 10. The metasurface-based camera attains an average polarization reconstruction error below 0.981%, markedly surpassing state-of-the-art commercial LWIR polarimetric systems. Additionally, the new camera presents excellent polarimetric imaging capability for complex scenes. To the best of our knowledge, this represents the world's first LWIR polarimetric camera utilizing the metasurface optical chip with performance superior to commercial cameras, promoting the practical development of metasurface-integrated devices.

physics.optics↗

arXiv2025

From Generic to Specialized: A Subspecialty Diagnostic System Powered by Self-Supervised Learning for Cervical Histopathology

Cervical cancer remains a major malignancy, necessitating extensive and complex histopathological assessments and comprehensive support tools. Although deep learning shows promise, these models still lack accuracy and generalizability. General foundation models offer a broader reach but remain limited in capturing subspecialty-specific features and task adaptability. We introduce the Cervical Subspecialty Pathology (CerS-Path) diagnostic system, developed through two synergistic pretraining stages: self-supervised learning on approximately 190 million tissue patches from 140,000 slides to build a cervical-specific feature extractor, and multimodal enhancement with 2.5 million image-text pairs, followed by integration with multiple downstream diagnostic functions. Supporting eight diagnostic functions, including rare cancer classification and multimodal Q&A, CerS-Path surpasses prior foundation models in scope and clinical applicability. Comprehensive evaluations demonstrate a significant advance in cervical pathology, with prospective testing on 3,173 cases across five centers maintaining 99.38% screening sensitivity and excellent generalizability, highlighting its potential for subspecialty diagnostic translation and cervical cancer screening.

cs.CV↗

arXiv2025

SDQ-LLM: Sigma-Delta Quantization for 1-bit LLMs of any size

Large language models (LLMs) face significant computational and memory challenges, making extremely low-bit quantization crucial for their efficient deployment. In this work, we introduce SDQ-LLM: Sigma-Delta Quantization for 1-bit LLMs of any size, a novel framework that enables extremely low-bit quantization of LLMs while preserving their linguistic reasoning capabilities. A distinctive feature of SDQ-LLM is the continuous adjustability of the Over-Sampling Ratio (OSR), enabling dynamic adaptation to memory or VRAM constraints by selecting fractional OSR (e.g. 2.5 times) for an optimal trade-off between model size and accuracy. SDQ-LLM uses upsampling combined with Sigma-Delta Quantizer to binarize or ternarize LLMs weights, encoding high-precision parameters into 1-bit or 1.58-bit representations, replacing the multiplication operations within linear layers with addition. This approach significantly enhances inference efficiency under extremely low-bit quantization. To further reduce the loss of quantization precision, we incorporate Hadamard-based weight smoothing prior to quantization, improving the stability and robustness of the weight representations. Furthermore, to fully leverage the continuity of the OSR and reduce precision loss, recognizing the correlation between quantization sensitivity and weight variance, we propose a fine-grained, layer- and linear-wise OSR allocation strategy, MultiOSR. This strategy distributes OSR both across layers and within each layer, based on weight variance and parameter scale. Finally, extensive experiments on OPT and LLaMA model families demonstrate that SDQ-LLM achieves a more efficient and high-precision performance even under highly aggressive low-OSR settings. Our code is available at https://github.com/Dreamlittlecat/LLM-Quant-Factory.

cs.LG↗

arXiv2025

Chain or tree? Re-evaluating complex reasoning from the perspective of a matrix of thought

Large Language Models (LLMs) face significant accuracy degradation due to insufficient reasoning ability when dealing with complex and abstract tasks. Thought structures such as Chain of Thought (CoT) and Tree of Thought (ToT) focus on enhancing the reasoning capability of LLMs. However, they suffer from inherent drawbacks such as redundancy within the same layer of the tree structure and the singularity of the paths in the chain structure. Some studies have utilized Retrieval-Augmented Generation (RAG) methods to enhance CoT and ToT in mitigating hallucinations in LLMs, yet the fundamental shortcomings of the thought structures still persist. Furthermore, when dealing with multi-entity and multi-hop information, the retrieved verification knowledge often contains large amounts of fragmented, superficial, or even erroneous data, misleading the reasoning process of LLMs. To address these issues, we propose the Matrix of Thought (MoT), a novel and efficient thought structure for LLMs. MoT explores problems in both horizontal and vertical dimensions through a "column-cell communication" mechanism, enabling LLMs to actively engage in multi-strategy and deep thinking while reducing redundancy in the thought nodes within the column cells, thereby enhancing the reasoning capability of LLMs. Additionally, through a fact-correction mechanism, it leverages the knowledge graph triples retrieved by RAG and the original text to construct knowledge units and correct erroneous answers. To validate the effectiveness of this method, we conducted extensive experiments in three tasks: 24-point game, question answering evaluation, and proposition writing.The results demonstrate that our framework outperforms state-of-the-art methods, with reasoning time only 14.4\% of that of the baseline method, proving its efficiency and accuracy. The code for framework is available at https://github.com/lyfiter/mtqa.

cs.CL↗

arXiv2025

UnifiedRL: A Reinforcement Learning Algorithm Tailored for Multi-Task Fusion in Large-Scale Recommender Systems

As the last pivotal stage of Recommender System (RS), Multi-Task Fusion (MTF) is responsible for combining multiple scores outputted by Multi-Task Learning (MTL) model into a final score to maximize user satisfaction. Recently, to optimize long-term user satisfaction, Reinforcement Learning (RL) is used for MTF in RSs. However, the existing offline RL algorithms used for MTF have the following severe problems: a) To avoid Out-of-Distribution (OOD), their constraints are overly strict, which seriously damage performance; b) They are unaware of the exploration policy used to collect training data, only suboptimal policy can be learned; c) Their exploration policies are inefficient and hurt user experience. To solve the above problems, we propose an innovative method called UnifiedRL tailored for MTF in large-scale RSs. UnifiedRL seamlessly integrates offline RL model with its custom exploration policy to relax overly strict constraints, which is different from existing RL-MTF methods and significantly improves performance. In addition, compared to existing exploration policies, UnifiedRL's custom exploration policy is highly efficient, enabling frequent online exploration and offline training iterations, which further improves performance. Extensive offline and online experiments are conducted in a large-scale RS. The results demonstrate that UnifiedRL outperforms other existing MTF methods remarkably, achieving a +4.64% increase in user valid consumption and a +1.74% increase in user duration time. To the best of our knowledge, UnifiedRL is the first RL algorithm tailored for MTF in RSs and has been successfully deployed in multiple large-scale RSs since June 2023, yielding significant benefits.

cs.IR↗

arXiv2025

MoTiC: Momentum Tightness and Contrast for Few-Shot Class-Incremental Learning

Few-Shot Class-Incremental Learning (FSCIL) must contend with the dual challenge of learning new classes from scarce samples while preserving old class knowledge. Existing methods use the frozen feature extractor and class-averaged prototypes to mitigate against catastrophic forgetting and overfitting. However, new-class prototypes suffer significant estimation bias due to extreme data scarcity, whereas base-class prototypes benefit from sufficient data. In this work, we theoretically demonstrate that aligning the new-class priors with old-class statistics via Bayesian analysis reduces variance and improves prototype accuracy. Furthermore, we propose large-scale contrastive learning to enforce cross-category feature tightness. To further enrich feature diversity and inject prior information for new-class prototypes, we integrate momentum self-supervision and virtual categories into the Momentum Tightness and Contrast framework (MoTiC), constructing a feature space with rich representations and enhanced interclass cohesion. Experiments on three FSCIL benchmarks produce state-of-the-art performances, particularly on the fine-grained task CUB-200, validating our method's ability to reduce estimation bias and improve incremental learning robustness.

cs.CV↗

arXiv2025

Dynamic User Interest Augmentation via Stream Clustering and Memory Networks in Large-Scale Recommender Systems

Recommender System (RS) provides personalized recommendation service based on user interest. However, lots of users' interests are sparse due to lacking consumption behaviors, making it challenging to provide accurate recommendations for them, which is widespread in large-scale RSs. In particular, efficiently solving this problem in the ranking stage of RS is an even greater challenge, which requires an end-to-end and real-time approach. To solve this problem, we propose an innovative method called Dynamic User Interest Augmentation (DUIA). DUIA enhances user interest including user profile and user history behavior sequences by generating enhancement vectors and personalized enhancement vectors through dynamic stream clustering of similar users and relevant items from multiple perspectives. To realize stream clustering, we specially design an algorithm called Gradient-based Hierarchical Clustering Algorithm (GHCA) for DUIA, which performs clustering via gradient descent and stores the cluster centers in memory networks. Extensive offline and online experiments demonstrate that DUIA not only significantly improves model performance for users with sparse interests but also delivers notable gains for other users. As an end-to-end method, DUIA can be easily integrated with existing models. Furthermore, DUIA is also used for long-tail items and cold-start problem, which also yields excellent improvements. Since 2022, DUIA has been successfully deployed in multiple industrial RSs in Tencent and was made public in May 2024. Moreover, the thoughts behind DUIA, dynamic stream clustering and similarity-based enhancement, have inspired relevant works and have also been applied in other stages of RS.

cs.IR↗

arXiv2025

Federated Hypergraph Learning with Local Differential Privacy: Toward Privacy-Aware Hypergraph Structure Completion

The rapid growth of graph-structured data necessitates partitioning and distributed storage across decentralized systems, driving the emergence of federated graph learning to collaboratively train Graph Neural Networks (GNNs) without compromising privacy. However, current methods exhibit limited performance when handling hypergraphs, which inherently represent complex high-order relationships beyond pairwise connections. Partitioning hypergraph structures across federated subsystems amplifies structural complexity, hindering high-order information mining and compromising local information integrity. To bridge the gap between hypergraph learning and federated systems, we develop FedHGL, a first-of-its-kind framework for federated hypergraph learning on disjoint and privacy-constrained hypergraph partitions. Beyond collaboratively training a comprehensive hypergraph neural network across multiple clients, FedHGL introduces a pre-propagation hyperedge completion mechanism to preserve high-order structural integrity within each client. This procedure leverages the federated central server to perform cross-client hypergraph convolution without exposing internal topological information, effectively mitigating the high-order information loss induced by subgraph partitioning. Furthermore, by incorporating two kinds of local differential privacy (LDP) mechanisms, we provide formal privacy guarantees for this process, ensuring that sensitive node features remain protected against inference attacks from potentially malicious servers or clients. Experimental results on seven real-world datasets confirm the effectiveness of our approach and demonstrate its performance advantages over traditional federated graph learning methods.

cs.LG↗

arXiv2025

Scavenger+: Revisiting Space-Time Tradeoffs in Key-Value Separated LSM-trees

Key-Value Stores (KVS) based on log-structured merge-trees (LSM-trees) are widely used in storage systems but face significant challenges, such as high write amplification caused by compaction. KV-separated LSM-trees address write amplification but introduce significant space amplification, a critical concern in cost-sensitive scenarios. Garbage collection (GC) can reduce space amplification, but existing strategies are often inefficient and fail to account for workload characteristics. Moreover, current key-value (KV) separated LSM-trees overlook the space amplification caused by the index LSM-tree. In this paper, we systematically analyze the sources of space amplification in KV-separated LSM-trees and propose Scavenger+, which achieves a better performance-space trade-off. Scavenger+ introduces (1) an I/O-efficient garbage collection scheme to reduce I/O overhead, (2) a space-aware compaction strategy based on compensated size to mitigate index-induced space amplification, and (3) a dynamic GC scheduler that adapts to system load to make better use of CPU and storage resources. Extensive experiments demonstrate that Scavenger+ significantly improves write performance and reduces space amplification compared to state-of-the-art KV-separated LSM-trees, including BlobDB, Titan, and TerarkDB.

cs.DB↗

arXiv2025

OpenCXD: An Open Real-Device-Guided Hybrid Evaluation Framework for CXL-SSDs

The advent of Compute Express Link (CXL) enables SSDs to participate in the memory hierarchy as large-capacity, byte-addressable memory devices. These CXL-enabled SSDs (CXL-SSDs) offer a promising new tier between DRAM and traditional storage, combining NAND flash density with memory-like access semantics. However, evaluating the performance of CXL-SSDs remains difficult due to the lack of hardware that natively supports the CXL.mem protocol on SSDs. As a result, most prior work relies on hybrid simulators combining CPU models augmented with CXL.mem semantics and SSD simulators that approximate internal flash behaviors. While effective for early-stage exploration, this approach cannot faithfully model firmware-level interactions and low-level storage dynamics critical to CXL-SSD performance. In this paper, we present OpenCXD, a real-device-guided hybrid evaluation framework that bridges the gap between simulation and hardware. OpenCXD integrates a cycle-accurate CXL.mem simulator on the host side with a physical OpenSSD platform running real firmware. This enables in-situ firmware execution triggered by simulated memory requests. Through these contributions, OpenCXD reflects device-level phenomena unobservable in simulation-only setups, providing critical insights for future firmware design tailored to CXL-SSDs.

cs.AR↗

arXiv2025

WEC-DG: Multi-Exposure Wavelet Correction Method Guided by Degradation Description

Multi-exposure correction technology is essential for restoring images affected by insufficient or excessive lighting, enhancing the visual experience by improving brightness, contrast, and detail richness. However, current multi-exposure correction methods often encounter challenges in addressing intra-class variability caused by diverse lighting conditions, shooting environments, and weather factors, particularly when processing images captured at a single exposure level. To enhance the adaptability of these models under complex imaging conditions, this paper proposes a Wavelet-based Exposure Correction method with Degradation Guidance (WEC-DG). Specifically, we introduce a degradation descriptor within the Exposure Consistency Alignment Module (ECAM) at both ends of the processing pipeline to ensure exposure consistency and achieve final alignment. This mechanism effectively addresses miscorrected exposure anomalies caused by existing methods' failure to recognize 'blurred' exposure degradation. Additionally, we investigate the light-detail decoupling properties of the wavelet transform to design the Exposure Restoration and Detail Reconstruction Module (EDRM), which processes low-frequency information related to exposure enhancement before utilizing high-frequency information as a prior guide for reconstructing spatial domain details. This serial processing strategy guarantees precise light correction and enhances detail recovery. Extensive experiments conducted on multiple public datasets demonstrate that the proposed method outperforms existing algorithms, achieving significant performance improvements and validating its effectiveness and practical applicability.

cs.CV↗

arXiv2025

KGV: Integrating Large Language Models with Knowledge Graphs for Cyber Threat Intelligence Credibility Assessment

Cyber threat intelligence (CTI) is a crucial tool to prevent sophisticated, organized, and weaponized cyber attacks. However, few studies have focused on the credibility assessment of CTI, and this work still requires manual analysis by cybersecurity experts. In this paper, we propose Knowledge Graph-based Verifier (KGV), the first framework integrating large language models (LLMs) with simple structured knowledge graphs (KGs) for automated CTI credibility assessment. Unlike entity-centric KGs, KGV constructs paragraph-level semantic graphs where nodes represent text segments connected through similarity analysis, which effectively enhances the semantic understanding ability of the model, reduces KG density and greatly improves response speed. Experimental results demonstrate that our KGV outperforms state-of-the-art fact reasoning methods on the CTI-200 dataset, achieving a 5.7\% improvement in F1. Additionally, it shows strong scalability on factual QA and fake news detection datasets. Compared to entity-based knowledge graphs (KGs) for equivalent-length texts, our structurally simple KG reduces node quantities by nearly two-thirds while boosting precision by 1.7\% and cutting response time by 46.7\%. In addition, we have created and publicly released the first CTI credibility assessment dataset, CTI-200. Distinct from CTI identification datasets, CTI-200 refines CTI summaries and key sentences to focus specifically on credibility assessment.

cs.CR↗

arXiv2025

Advancing Seasonal Prediction of Tropical Cyclone Activity with a Hybrid AI-Physics Climate Model

Machine learning (ML) models are successful with weather forecasting and have shown progress in climate simulations, yet leveraging them for useful climate predictions needs exploration. Here we show this feasibility using Neural General Circulation Model (NeuralGCM), a hybrid ML-physics atmospheric model developed by Google, for seasonal predictions of large-scale atmospheric variability and Northern Hemisphere tropical cyclone (TC) activity. Inspired by physical model studies, we simplify boundary conditions, assuming sea surface temperature (SST) and sea ice follow their climatological cycle but persist anomalies present at the initialization time. With such forcings, NeuralGCM can generate 100 simulation days in ~8 minutes with a single Graphics Processing Unit (GPU), while simulating realistic atmospheric circulation and TC climatology patterns. This configuration yields useful seasonal predictions (July to November) for the tropical atmosphere and various TC activity metrics. Notably, the predicted and observed TC frequency in the North Atlantic and East Pacific basins are significantly correlated during 1990 to 2023 (r=~0.7), suggesting prediction skill comparable to existing physical GCMs. Despite challenges associated with model resolution and simplified boundary forcings, the model-predicted interannual variations demonstrate significant correlations with the observation, including the sub-basin TC tracks (p<0.1) and basin-wide accumulated cyclone energy (p<0.01) of the North Atlantic and North Pacific basins. These findings highlight the promise of leveraging ML models with physical insights to model TC risks and deliver seamless weather-climate predictions.

physics.ao-ph↗

arXiv2025

LRCTI: A Large Language Model-Based Framework for Multi-Step Evidence Retrieval and Reasoning in Cyber Threat Intelligence Credibility Verification

Verifying the credibility of Cyber Threat Intelligence (CTI) is essential for reliable cybersecurity defense. However, traditional approaches typically treat this task as a static classification problem, relying on handcrafted features or isolated deep learning models. These methods often lack the robustness needed to handle incomplete, heterogeneous, or noisy intelligence, and they provide limited transparency in decision-making-factors that reduce their effectiveness in real-world threat environments. To address these limitations, we propose LRCTI, a Large Language Model (LLM)-based framework designed for multi-step CTI credibility verification. The framework first employs a text summarization module to distill complex intelligence reports into concise and actionable threat claims. It then uses an adaptive multi-step evidence retrieval mechanism that iteratively identifies and refines supporting information from a CTI-specific corpus, guided by LLM feedback. Finally, a prompt-based Natural Language Inference (NLI) module is applied to evaluate the credibility of each claim while generating interpretable justifications for the classification outcome. Experiments conducted on two benchmark datasets, CTI-200 and PolitiFact show that LRCTI improves F1-Macro and F1-Micro scores by over 5%, reaching 90.9% and 93.6%, respectively, compared to state-of-the-art baselines. These results demonstrate that LRCTI effectively addresses the core limitations of prior methods, offering a scalable, accurate, and explainable solution for automated CTI credibility verification

cs.CR↗

arXiv2025

Parallax QAMA: Novel Downlink Multiple Access for MISO Systems with Simple Receivers

In this paper, we propose a novel downlink multiple access system with a multi-antenna transmitter and two single-antenna receivers, inspired by the underlying principles of hierarchical quadrature amplitude modulation (H-QAM) based multiple access (QAMA) and space-division multiple access (SDMA). In the proposed scheme, coded bits from two users are split and assigned to one shared symbol and two private symbols carried by different beams. Based on joint symbol mapping of H-QAM constellations and phase-aligned precoding at the transmitter, each receiver observes a different H-QAM constellation with Gray mapping, a unique parallax feature not shared by existing schemes. In addition to avoiding successive interference cancellation (SIC), each user independently demodulates its own bits on separate I and Q branches with calculations based on closed-form expressions. Hence the receiver complexity is on par with that of orthogonal multiple access (OMA), which is much lower than that in other competing alternatives such as non-orthogonal multiple access (NOMA) and rate-splitting multiple access (RSMA). We carry out system optimization and determine the achievable rate region. Numerical results show that the proposed system has a larger rate region relative to other benchmark schemes with receivers not using SIC, and even achieves a comparable rate region to those benchmark schemes with SIC receivers.

cs.IT↗

arXiv2025

ReF-LLE: Personalized Low-Light Enhancement via Reference-Guided Deep Reinforcement Learning

Low-light image enhancement presents two primary challenges: 1) Significant variations in low-light images across different conditions, and 2) Enhancement levels influenced by subjective preferences and user intent. To address these issues, we propose ReF-LLE, a novel personalized low-light image enhancement method that operates in the Fourier frequency domain and incorporates deep reinforcement learning. ReF-LLE is the first to integrate deep reinforcement learning into this domain. During training, a zero-reference image evaluation strategy is introduced to score enhanced images, providing reward signals that guide the model to handle varying degrees of low-light conditions effectively. In the inference phase, ReF-LLE employs a personalized adaptive iterative strategy, guided by the zero-frequency component in the Fourier domain, which represents the overall illumination level. This strategy enables the model to adaptively adjust low-light images to align with the illumination distribution of a user-provided reference image, ensuring personalized enhancement results. Extensive experiments on benchmark datasets demonstrate that ReF-LLE outperforms state-of-the-art methods, achieving superior perceptual quality and adaptability in personalized low-light image enhancement.

cs.CV↗

arXiv2025

A fluctuation-dissipation theorem perspective on radiative responses to temperature perturbations

Radiative forcing drives warming in the Earth system, leading to changes in sea surface temperatures (SSTs) and associated radiative feedbacks. The link between changes in the top-of-the-atmosphere (TOA) net radiative flux and SST patterns, known as the "pattern effect", is typically diagnosed by studying the response of atmosphere-only models to SST perturbations. In this work, we diagnose the pattern effect through response theory, by performing idealized warming perturbation experiments from unperturbed data alone. First, by studying the response at short time scales, where the response is dominated by atmospheric variability, we recover results that agree with the literature. Second, by extending the framework to longer time scales, we capture coupled interactions between the slow ocean component and the atmosphere, yielding a novel "sensitivity map" quantifying the response of the net radiative flux to SST perturbations in the coupled system. Here, feedbacks are captured by a spatiotemporal response operator, rather than time-independent maps as in traditional studies. Both formulations skillfully reconstruct changes in externally forced simulations and provide practical strategies for climate studies. The key distinction lies in their perspectives on climate feedbacks. The first formulation, closely aligned with prediction tasks, follows the traditional view in which slow variables, such as SSTs, exert a one-way influence on fast variables. The second formulation broadens this perspective by incorporating spatiotemporal interactions across state variables. This alternative approach explores how localized SST perturbations can alter the coupled dynamics, leading to temperature changes in remote areas and further impacting the radiative fluxes at later times.

physics.ao-ph↗

arXiv2025

Adapting Network Information into Semantics for Generalizable and Plug-and-Play Multi-Scenario Network Diagnosis

Leverage large language model (LLM) to refer the fault is considered to be a potential solution for intelligent network fault diagnosis. However, how to represent network information in a paradigm that can be understood by LLMs has always been a core issue that has puzzled scholars in the field of network intelligence. To address this issue, we propose LLM-based Network Semantic Generation (LNSG) algorithm, which integrates semanticization and symbolization methods to uniformly describe the entire multi-modal network information. Based on the LNSG and LLMs, we present NetSemantic, a plug-and-play, data-independent, network information semantic fault diagnosis framework. It enables rapid adaptation to various network environments and provides efficient fault diagnosis capabilities. Experimental results demonstrate that NetSemantic excels in network fault diagnosis across various complex scenarios in a zero-shot manner.

cs.NI↗