arXiv subjects

Ming Zhao

Publications and source records attributed to Ming Zhao.

At least 19 recordsLinked to original sources

arXiv2026

NGM: A Plug-and-Play Training-Free Memory Module for LLMs

Recent studies have introduced memory-augmented architectures that decouple knowledge storage from neural computation, enabling more direct access to stored representations. However, existing approaches often rely on learned memory embeddings or explicit memory and retrieval components, requiring additional training or increasing architectural complexity. To address these limitations, we propose N-gram Memory (NGM), a training-free, plug-and-play module that repurposes a backbone model's pretrained token embedding space as memory. NGM consists of a Causal N-Gram Encoder and a Cosine-Gated Memory Injector. The former constructs N-gram representations by directly aggregating pretrained token embeddings, while the latter uses cosine similarity with ReLU gating to adaptively control the amount of retrieved memory injected into contextual representations. Unlike conventional memory-augmented architectures, NGM requires neither additional embedding training nor a separately learned memory bank or external retrieval infrastructure. We evaluate NGM on ten benchmarks across six language-model backbones spanning two model families and multiple scales, including four Qwen3 models (0.6B, 1.7B, 4B, and 8B) and two Llama-3.2-Instruct models (1B and 3B). NGM consistently improves the ten-task average across all six backbones. Notably, it improves HumanEval by 4.26, HumanEval+ by 2.41, and GPQA-Diamond by 6.56 percentage points. Furthermore, NGM extends beyond text-only language models to vision-language models, with Qwen3-VL-2B achieving improvements across all five VLM benchmarks. Project page: https://pioneerqyw.github.io/ngm-project/

cs.AI↗

arXiv2026

Graph Topology Meets Energy-Driven Pathfinding: Preserving Contextual Consistency in Retrieval-Augmented Generation

Retrieval-augmented generation (RAG) frequently integrates knowledge graphs (KGs) for complex multi-hop question answering. However, their performance is increasingly bottlenecked by the knowledge structuring and reasoning processes. (i) Severe semantic loss: Distilling continuous texts into discrete symbols strips away contextual nuances and narrative coherence. (ii) Pronounced modality mismatch: Fragmented graph structures clash with the continuous texts LLMs are pre-trained on, while rewriting them into natural language introduces generative noise and omits information. To address these limitations, we propose ThreadRAG, a semantic-topology-based RAG framework for context-preserving knowledge structuring and modality-native evidence reasoning. ThreadRAG introduces a Semantic Topology Mapper (STM), which organizes retrieved documents into context-intact Semantic Nodes and constructs a Semantic Topology Graph (STG) through narrative, semantic, and lexical edges. It further incorporates Energy-Driven Pathfinding (EDP), which identifies query-relevant subgraphs via energy diffusion and deterministically linearizes them into coherent textual evidence paths. Experiments on three multi-hop QA benchmarks demonstrate that ThreadRAG consistently outperforms strong baselines. Notably, it achieves a 14.4% EM improvement over text-centric methods on HotpotQA and surpasses the best baseline by 5.8% EM on 2WikiMultihop, while reducing token cost to 23.3% of the strongest baseline.

cs.CL↗

arXiv2026

Every Batch Is Its Own Validation Set: Leave-One-Out Gradient Matching for Online Data Selection in LLM Fine-Tuning

Online batch selection fine-tunes a language model on the most useful part of each candidate batch. Selectors that match the gradient of the candidate batch are attractive because they need no held-out data, yet they rarely beat training on the whole batch. We show why. In-sample gradient matching uses every example as part of its own target, so its objective credits each example with its own gradient noise. This is the covariance penalty that makes training error optimistic, now sitting on the diagonal of the gradient Gram matrix: it steers selection toward the noisiest examples and makes the full batch the best solution the objective can reach. The fix costs nothing. For each example, the other candidates form an independent sample of the data distribution, so removing the diagonal turns the matching objective into an unbiased estimate of the update's error with respect to the population gradient. The minimizer of this leave-one-out objective weights examples by their gradient signal-to-noise ratio (SNR), and whenever per-example SNR is heterogeneous enough, half of a batch yields a lower-error update than the whole batch; we give the exact condition. We build \method{} on this principle. It computes the Gram matrix in the metric of the Adam preconditioner during the ordinary backward pass, selects a weighted subset greedily with a $(1-e^{-γ})$ guarantee, and uses no held-out data. Across four fine-tuning tasks and seven backbones from 1.5B to 8B parameters, LOOM improves on full-batch training by 2.3 and 2.4 points on Llama-3.1-8B and Qwen2.5-7B, exceeds every in-sample gradient matcher by 2.4 points and the validation-guided GREATS and OPUS by 1.6--2.0, and selects injected label noise at under a fifth of its base rate.

cs.LG↗

arXiv2026

Information-Entropy-Driven Fault Propagation Modeling for Probabilistic Network Performance Prediction

Network faults can trigger cascading effects that cause abrupt and nonstationary performance degradation. Existing learning-based performance predictors mainly focus on normal operation or treat fault-induced topology and routing changes as static inputs, and typically produce deterministic point estimates. They overlook fault-propagation dynamics and uncertainty in performance evolution. The predefined-rule and purely data-driven propagation models lack a unified representation of fault definition, propagation mechanism, and impact quantification. Additionally, generic denoisers in conditional diffusion models fail to incorporate fault propagation into uncertainty modeling. To address these limitations, we propose an information-entropy-driven fault propagation paradigm (IEFP) that characterizes fault propagation via relative entropy, mutual information and transfer entropy. We then design a fault-aware graph message-passing mechanism that propagation contexts modulate network representation learning. We further develop FEMNet, which employs this mechanism as a tailored denoiser within a conditional diffusion model to enable probabilistic network performance prediction under complex fault scenarios. Compared with the strongest baselines, IEFP improves fault-prediction performance, while FEMNet reduces errors in both point and probabilistic KPI prediction.

cs.NI↗

arXiv2026

An evidence-guided reinforcement learning method to improve psychiatric reasoning in small language models

Privacy and computational constraints limit the use of large language models in psychiatry, while adapting small language models (SLMs) often requires substantial data and expert annotation. We developed ClinMPO, an evidence-guided reinforcement-learning framework guided by the psychiatrist-defined Clinical Psychiatry Thinking Strategy (CPTS). ClinMPO uses ClinRM, a reward model trained on 18,569 question--answer pairs from 4,474 psychiatry articles. We evaluated four Qwen3 sizes on 1,737 model-screened questions. ClinMPO outperformed Base, supervised fine-tuning and standard group relative policy optimization across scales. From responses by 300 senior pre-licensure medical students, we established the human baseline, a medical-student reference. The 4B model approached this baseline, whereas the 8B model surpassed it and ranked first among 31 models and post-training variants. ClinMPO improved performance across two complementary schemes covering ICD-11 diagnostic categories and psychiatric practice competencies. Blinded assessment by three clinicians showed improved rationale quality across CPTS criteria. These findings highlight how existing clinical evidence and specialist knowledge can be incorporated into the development of medical AI systems through evidence-guided learning.

cs.CL↗

arXiv2026

SPFR: Semantic Potential Field Routing for the Distributed Internet of Agents

In a distributed Internet of Agents (IoA) without centralized routing control, routing tasks to capability-matched executors is challenging because destinations are not predetermined and agents have bounded local service views. Discover-then-forward approaches, by contrast, select an executor before network forwarding and therefore do not directly support reselection when additional candidates become visible downstream. We introduce Semantic Potential Field Routing (SPFR), a distributed IoA routing algorithm that integrates executor discovery and reselection into hop-by-hop forwarding. SPFR represents each executor visible in a local semantic forwarding information base (FIB) as a task-conditioned semantic potential source, with utility setting its strength and hop distance inducing exponential attenuation. At each hop, the forwarding agent recomputes these potentials, reselects the dominant executor, and forwards the task one hop toward it. Under task-consistent frozen-FIB conditions, we prove loop freedom and finite-hop termination and derive an explicit additive error bound under bounded visibility relative to the full-visibility objective. Extensive simulations on real-world topologies show that SPFR approaches the realized utility of distributed utility-greedy routing and request-triggered global discovery while using fewer forwarding hops and substantially fewer request-triggered messages, and remains robust under network and service dynamics.

cs.NI↗

arXiv2026

EdgeFaaS: A Function-based Framework for Edge Computing

Edge computing brings unique challenges as the resources on the edge are highly diverse in capabilities and capacities, and highly distributed across many users and the physical world. Existing distributed computing frameworks cannot adequately handle this level of heterogeneity and distribution. This paper proposes EdgeFaaS, a novel function-based edge computing framework to enable edge applications to effectively utilize heterogeneous resources distributed across the Internet of Things (IoT), edge, and cloud for computing. It proposes function virtualization and storage virtualization to abstract distributed and heterogeneous physical resources and provides consistent virtual interfaces for deploying and executing functions and storing and accessing data. EdgeFaaS provides comprehensive support to diverse edge computing workflows, and at the same time allows users to flexibly adjust the configurations and explore various important tradeoffs. To demonstrate its usability, the paper also presents the implementation and evaluation of three representative workflows on EdgeFaaS for video analytics, federated learning, and audio classification, on a real testbed of 100+ geographically distributed IoT devices, edge servers, and cloud services. EdgeFaaS allows users to flexibly explore the deployment configurations of these workflows over distributed and heterogeneous resources. For example, users can easily vary the function placement of the video processing pipeline across IoT, edge, and cloud resources and study the tradeoff between computation and communication costs; users can also flexibly adjust the cluster count and size in the hierarchical federated learning system and explore the tradeoff between training accuracy and speed.

cs.DC↗

arXiv2026

Co-Adaptive Multi-Task LoRA: Transfer-Aware, Label-Free Control of Domain Participation

Fine-tuning a single low-rank adapter on many domains at once is multi-task learning: the domains must be co-learned, and how they share the adapter decides whether they help or hurt one another. Most efficient fine-tuning pipelines ignore this and train on a fixed, uniform mixture, leaving two coupled questions unanswered: how much should each domain participate, and which domains should be co-trained given that some transfer positively and others interfere? We show that both answers can be read off cheaply and without labels. A forward pass of the current shared adapter over a small unlabeled probe yields, per domain, a competence signal whose level tracks remaining headroom and whose trajectory tracks learning speed; the drift of these probe representations yields a signed cross-domain affinity that predicts pairwise transfer. We fold both into CoDA, a co-adaptive controller that solves a small entropy-regularized quadratic program on the simplex to set each domain's participation -- jointly its loss weight and its share of the sampled data -- rewarding high-headroom, still-learning, mutually synergistic domains and damping interfering ones. The controller is forward-only, adds no trainable parameters, and wraps any multi-task LoRA pipeline. Across five heterogeneous domains and two backbones, CoDA improves the average over uniform mixing, learned mixtures, gradient-surgery multi-task optimizers, and online data selection while using half the data, and lowers cross-domain gradient conflict. We prove that the competence signal tracks domain risk, that the participation program has a unique fixed point reached by a contraction, and that its solution performs transfer-aware water-filling; analysis, ablations, and controls corroborate each claim.

cs.LG↗

arXiv2026

Predicting Mergeability of Parameter-Efficient Fine-Tuning Updates

Low-rank adaptation (LoRA) makes it cheap to train many domain- and task-specific language model adapters, but whether two adapters can be merged is usually discovered only after both have been fully trained and evaluated. This late feedback is costly: adapters that are strong in isolation can interfere destructively once their updates are combined. We ask whether this outcome can be anticipated. We formalize adapter mergeability as the degree to which an adapter preserves its single-task utility after merging, and show that it can be forecast from signals measured in the first few percent of training -- chiefly how the low-rank updates and their gradients align across tasks and how much they disturb shared representations. We package these signals into MergeProbe, a lightweight predictor that estimates pairwise and set-level retention and turns the estimate into a concrete decision: merge directly, reweight, prune, or route. On MERGE-PEFT, a five-domain benchmark spanning math, code, science, instruction following, and safety, MergeProbe attains the best average and worst-case retention among strong interference-aware merge baselines while adding far less deployment overhead than full task routing. This turns LoRA merging from a post-hoc engineering step into an anticipatory measurement problem.

cs.LG↗

arXiv2026

PropLLM: Propagation-Aware Scene Reconstruction for Network Fault Diagnosis

Network faults propagate layer by layer along topology and protocol dependencies, yet operations systems typically observe only symptomatic alerts at the tail end of propagation chains, where distinct root-cause faults may produce highly similar end-point symptoms. Existing approaches, whether rule-based, machine learning (ML)-based, or large language model (LLM)-based, fundamentally map the alert set to a diagnosis in a single pass and are structurally incapable of resolving this end-point ambiguity. This paper proposes PropLLM, which is the first to integrate the hop-by-hop scene reconstruction paradigm with the generative reasoning capabilities of LLMs. Starting from end-point alerts, PropLLM traces back hop-by-hop along the propagation path, retrieving verifiable factual evidence from a dual-layer knowledge graph (KG) at each hop, while the proposed Temporal Causal Propagation Attention (TCPA) mechanism encodes known topological causal priors directly into the attention computation to guide the model along the correct causal direction, ultimately localizing the root cause and determining the fault type through a fully evidenced causal chain. On a real-world Wi-Fi multimodal fault dataset, PropLLM improves fault type diagnosis accuracy by 3.9\% and root cause localization accuracy by 4.7\% over the strongest baseline, while reducing the hallucination rate by 50.8\%. Supplementary experiments on the TeleLogs 5G dataset further demonstrate the effectiveness of the proposed method across different network scenarios.

cs.AI↗

arXiv2026

Toward Realistic Wi-Fi Fault Diagnosis: A Multi-Modal Benchmark

Intelligent network operation and maintenance systems in modern networks continuously generate large volumes of multi-modal operational data. However, Wi-Fi fault diagnosis under heterogeneous operational environments remains insufficiently understood. We build a real-world Wi-Fi testbed deployed in campus working environments with an automated fault injection system, and collect a multi-modal Wi-Fi fault dataset containing over 10,000 fault samples across diverse wireless scenarios. To the best of our knowledge, this is among the first publicly available datasets jointly capturing heterogeneous cross-layer operational observations for Wi-Fi fault diagnosis. Based on this dataset, we establish a unified benchmark spanning multiple diagnosis tasks, operational modalities, and representative diagnosis paradigms. Experimental results indicate that effectively leveraging heterogeneous operational data remains challenging for existing diagnosis approaches. We further evaluate emerging LLM-based approaches and develop a reasoningoriented evaluation framework to assess the consistency between generated diagnostic analyses and actual network conditions. Our findings suggest several important considerations for future multi-modal Wi-Fi diagnosis.

cs.NI↗

arXiv2026

Ivy-Fake: A Unified Explainable Framework and Benchmark for Image and Video AIGC Detection

The rapid development of Artificial Intelligence Generated Content (AIGC) techniques has enabled the creation of high-quality synthetic content, but it also raises significant security concerns. Current detection methods face two major limitations: (1) the lack of multidimensional explainable datasets for generated images and videos. Existing open-source datasets (e.g., WildFake, GenVideo) rely on oversimplified binary annotations, which restrict the explainability and trustworthiness of trained detectors. (2) Prior MLLM-based forgery detectors (e.g., FakeVLM) exhibit insufficiently fine-grained interpretability in their step-by-step reasoning, which hinders reliable localization and explanation. To address these challenges, we introduce Ivy-Fake, the first large-scale multimodal benchmark for explainable AIGC detection. It consists of over 106K richly annotated training samples (images and videos) and 5,000 manually verified evaluation examples, sourced from multiple generative models and real world datasets through a carefully designed pipeline to ensure both diversity and quality. Furthermore, we propose Ivy-xDetector, a reinforcement learning model based on Group Relative Policy Optimization (GRPO), capable of producing explainable reasoning chains and achieving robust performance across multiple synthetic content detection benchmarks. Extensive experiments demonstrate the superiority of our dataset and confirm the effectiveness of our approach. Notably, our method improves performance on GenImage from 86.88% to 96.32%, surpassing prior state-of-the-art methods by a clear margin.

cs.CV↗

arXiv2026

RadTwin: Generalizable Wireless Digital Twin for Dynamic Environments

Precisely modeling radio propagation in dynamic wireless environments is fundamental to the realization of wireless digital twins. Traditional ray tracing methods rely on accurate 3D models with detailed environment parameters, while recent neural radiance field approaches learn representations tied to specific static scenes, requiring retraining when environments change. In this paper, we propose RadTwin, a generalizable wireless digital twin framework that explicitly conditions on scene geometry, enabling adaptation to dynamic environments without retraining. RadTwin comprises three key components: 1) a scenario representation network that extracts high-level latent scene features from point clouds, 2) an electromagnetic ray tracing module that computes physics-informed sparse attention masks identifying voxels that physically contribute signals toward each query direction, and 3) a neural propagation decoder that aggregates relevant scene features through masked cross-attention to learn how radio propagation behaves within the given scene geometry. We evaluate RadTwin on a customized dataset of indoor scenes with varying furniture arrangements. Experimental results show that RadTwin achieves 31.6% higher SSIM (0.846 vs. 0.643) and 91.96% lower LPIPS (0.023 vs. 0.286) compared to NeRF2. RadTwin further demonstrates superior cross-scale performance and high generalization and data efficiency, representing a significant advancement toward practical digital network twins for dynamic wireless environments.

cs.NI↗

arXiv2026

MSADM: Large Language Model (LLM) Assisted End-to-End Network Health Management Based on Multi-Scale Semanticization

Network device and system health management is the foundation of modern network operations and maintenance. Traditional health management methods, relying on expert identification or simple rule-based algorithms, struggle to cope with the heterogeneous networks (HNs) environment. Moreover, current state-of-the-art distributed fault diagnosis methods, which utilize specific machine learning techniques, lack multi-scale adaptivity for heterogeneous device information, resulting in unsatisfactory diagnostic accuracy for HNs. In this paper, we develop an LLM-assisted end-to-end intelligent network health management framework. The framework first proposes a multi-scale data scaling method based on unsupervised learning to address the multi-scale data problem in HNs. Secondly, we combine the semantic rule tree with the attention mechanism to propose a Multi-Scale Semanticized Anomaly Detection Model (MSADM) that generates network semantic information while detecting anomalies. Finally, we embed a chain-of-thought-based large-scale language model downstream to adaptively analyze the fault diagnosis results and create an analysis report containing detailed fault information and optimization strategies. We compare our scheme with other fault diagnosis models and demonstrate that it performs well on several metrics of network fault diagnosis.

cs.NI↗

arXiv2026

SpineBench: A Clinically Salient, Level-Aware Benchmark Powered by the SpineMed-450k Corpus

Spine disorders affect 619 million people globally and are a leading cause of disability, yet AI-assisted diagnosis remains limited by the lack of level-aware, multimodal datasets. Clinical decision-making for spine disorders requires sophisticated reasoning across X-ray, CT, and MRI at specific vertebral levels. However, progress has been constrained by the absence of traceable, clinically-grounded instruction data and standardized, spine-specific benchmarks. To address this, we introduce SpineMed, an ecosystem co-designed with practicing spine surgeons. It features SpineMed-450k, the first large-scale dataset explicitly designed for vertebral-level reasoning across imaging modalities with over 450,000 instruction instances, and SpineBench, a clinically-grounded evaluation framework. SpineMed-450k is curated from diverse sources, including textbooks, guidelines, open datasets, and ~1,000 de-identified hospital cases, using a clinician-in-the-loop pipeline with a two-stage LLM generation method (draft and revision) to ensure high-quality, traceable data for question-answering, multi-turn consultations, and report generation. SpineBench evaluates models on clinically salient axes, including level identification, pathology assessment, and surgical planning. Our comprehensive evaluation of several recently advanced large vision-language models (LVLMs) on SpineBench reveals systematic weaknesses in fine-grained, level-specific reasoning. In contrast, our model fine-tuned on SpineMed-450k demonstrates consistent and significant improvements across all tasks. Clinician assessments confirm the diagnostic clarity and practical utility of our model's outputs.

cs.CV↗

arXiv2026

oneTwin: Online Digital Network Twin via Neural Radio Radiance Field

Digital network twin is a promising technology that replicates real-world networks in real-time and assists with the design, operation, and management of next-generation networks. However, existing approaches (e.g., simulator-based and neural-based) cannot effectively realize the digital network twin, in terms of fidelity, synchronicity, and tractability. In this paper, we propose oneTwin, the first online digital twin system, for the prediction of physical layer metrics. We architect the oneTwin system with two primary components: an enhanced simulator and a neural radio radiance field (NRRF). On the one hand, we achieve the enhanced simulator by designing a material tuning algorithm that incrementally optimizes the building materials to minimize the twin-to-real gap. On the other hand, we achieve the NRRF by designing a neural learning algorithm that continually updates its DNNs based on both online and simulated data from the enhanced simulator. We implement oneTwin system using Sionna RT as the simulator and developing new DNNs as the NRRF, under a public cellular network. Extensive experimental results show that, compared to state-of-the-art solutions, oneTwin achieves real-time updating (0.98s), with 36.39% and 57.50% reductions of twin-to-real gap under in-distribution and out-of-distribution test datasets, respectively.

cs.NI↗

arXiv2026

inRAN: Interpretable Online Bayesian Learning for Network Automation in Open Radio Access Networks

Emerging AI/ML techniques have been showing great potential in automating network control in open radio access networks (Open RAN). However, existing approaches heavily rely on blackbox policies parameterized by deep neural networks, which inherently lack interpretability, explainability, and transparency, and create substantial obstacles in practical network deployment. In this paper, we propose inRAN, a novel interpretable online Bayesian learning framework for network automation in Open RAN. The core idea is to integrate interpretable surrogate models and safe optimization solvers to continually optimize control actions, while adapting to non-stationary dynamics in real-world networks. We achieve the inRAN framework with three key components: 1) an interpretable surrogate model via ensembling Kolmogorov-Arnold Networks (KANs); 2) safe optimization solvers via integrating genetic search and trust-region descent method; 3) an online dynamics tracker via continual model learning and adaptive threshold offset. We implement inRAN in an end-to-end O-RAN-compliant network testbed, and conduct extensive over-the-air experiments with the focused use case of network slicing. The results show that, inRAN substantially outperforms state-of-the-art works, by guaranteeing the chance-based constraint with a 92.67% assurance ratio with comparative resource usage throughout the online network control, under unforeseeable time-evolving network dynamics.

cs.NI↗

arXiv2025

EnhancedRL: An Enhanced-State Reinforcement Learning Algorithm for Multi-Task Fusion in Recommender Systems

As a key stage of Recommender Systems (RSs), Multi-Task Fusion (MTF) is responsible for merging multiple scores output by Multi-Task Learning (MTL) into a single score, finally determining the recommendation results. Recently, Reinforcement Learning (RL) has been applied to MTF to maximize long-term user satisfaction within a recommendation session. However, due to limitations in modeling paradigm, all existing RL algorithms for MTF can only utilize user features and statistical features as the state to generate actions at the user level, but unable to leverage item features and other valuable features, which leads to suboptimal performance. Overcoming this problem requires a breakthrough in the existing modeling paradigm, yet, to date, no prior work has addressed it. To tackle this challenge, we propose EnhancedRL, an innovative RL algorithm. Unlike existing RL-MTF methods, EnhancedRL takes the enhanced state as input, incorporating not only user features but also item features and other valuable information. Furthermore, it introduces a tailored actor-critic framework - including redesigned actor and critics and a novel learning procedure - to optimize long-term rewards at the user-item pair level within a recommendation session. Extensive offline and online experiments are conducted in an industrial RS and the results demonstrate that EnhancedRL outperforms other methods remarkably, achieving a +3.84% increase in user valid consumption and a +0.58% increase in user duration time. To the best of our knowledge, EnhancedRL is the first work to address this challenge, and it has been fully deployed in a large-scale RS since September 14, 2023, yielding significant improvements.

cs.IR↗