arXiv subjects

Ziyuan Gao

Publications and source records attributed to Ziyuan Gao.

At least 19 recordsLinked to original sources

arXiv2026

MedCRP-CL: Continual Medical Image Segmentation via Bayesian Nonparametric Semantic Modality Discovery

Medical image segmentation faces a fundamental challenge in continual learning: data arrives sequentially from heterogeneous sources, yet effective continual learning requires discovering which tasks share sufficient structure to benefit from joint learning. Existing methods either apply uniform constraints across all tasks, causing catastrophic forgetting when tasks conflict, or require predefined task groupings that cannot anticipate future task diversity. We introduce MedCRP-CL, a framework that performs online task structure discovery and structure-aware continual learning. Leveraging the Chinese Restaurant Process (CRP), our method dynamically infers task groupings from clinical text prompts as tasks arrive, without requiring predefined cluster counts or access to future tasks. We term these discovered groupings semantic modalities, as they capture finer-grained structure than physical imaging modalities by integrating anatomical region and pathological context. Guided by this discovered structure, we maintain semantic modality-specific LoRA adapters regularized by intra-modality EWC, ensuring parameter isolation across dissimilar task groups while facilitating knowledge transfer within similar ones. The framework is also replay-free, storing only aggregate statistics rather than raw patient data. Experiments on 16 medical segmentation tasks across four imaging modalities demonstrate that MedCRP-CL achieves 73.3% Dice score with only 4.1% forgetting, outperforming the best baseline by 8.0% while requiring 6$\times$ fewer parameters. Code is available at https://github.com/zygao930/MedCRP-CL.

cs.CV↗

arXiv2025

AGENet: Adaptive Edge-aware Geodesic Distance Learning for Few-Shot Medical Image Segmentation

Medical image segmentation requires large annotated datasets, creating a significant bottleneck for clinical applications. While few-shot segmentation methods can learn from minimal examples, existing approaches demonstrate suboptimal performance in precise boundary delineation for medical images, particularly when anatomically similar regions appear without sufficient spatial context. We propose AGENet (Adaptive Geodesic Edge-aware Network), a novel framework that incorporates spatial relationships through edge-aware geodesic distance learning. Our key insight is that medical structures follow predictable geometric patterns that can guide prototype extraction even with limited training data. Unlike methods relying on complex architectural components or heavy neural networks, our approach leverages computationally lightweight geometric modeling. The framework combines three main components: (1) An edge-aware geodesic distance learning module that respects anatomical boundaries through iterative Fast Marching refinement, (2) adaptive prototype extraction that captures both global structure and local boundary details via spatially-weighted aggregation, and (3) adaptive parameter learning that automatically adjusts to different organ characteristics. Extensive experiments across diverse medical imaging datasets demonstrate improvements over state-of-the-art methods. Notably, our method reduces boundary errors compared to existing approaches while maintaining computational efficiency, making it highly suitable for clinical applications requiring precise segmentation with limited annotated data.

cs.CV↗

arXiv2025

Prompt-Aware Adaptive Elastic Weight Consolidation for Continual Learning in Medical Vision-Language Models

Medical AI systems face catastrophic forgetting when deployed in clinical settings, where models must learn new imaging protocols while retaining prior diagnostic capabilities. This challenge is particularly acute for medical vision-language models that must preserve complex cross-modal alignments between medical images and clinical terminology across diverse imaging modalities. We introduce Prompt- Aware Adaptive Elastic Weight Consolidation (PA-EWC), a novel continual learning approach that addresses catastrophic forgetting through prompt-guided parameter specialization. Our method systematically categorizes model parameters based on their functional roles in processing visual-descriptive, spatial-guided, and medical-semantic information, enabling targeted protection of critical knowledge while allowing adaptation to new clinical requirements. PA-EWC incorporates adaptive Fisher Information computation with gradient stability analysis and develops weighted complexity metrics based on medical terminology density. We evaluate our approach across five medical imaging datasets (Kvasir-SEG, ISIC 2018, CheXlocalize, BUSI, CAMUS) representing diverse modalities including endoscopy, dermoscopy, radiography, and ultrasound. Experimental results demonstrate that PA-EWC reduces catastrophic forgetting by up to 17.58% compared to baseline methods, with performance improvements of 4.30% on chest X-ray pathology localization and 6.06% on polyp segmentation.

cs.MM↗

arXiv2025

Reorienting Age-Friendly Frameworks for Rural Contexts: A Spatial Competence-Press Framework for Aging in Chinese Villages

While frameworks such as the WHO Age-Friendly Cities have advanced urban aging policy, rural contexts demand fundamentally different analytical approaches. The spatial dispersion, terrain variability, and agricultural labor dependencies that characterize rural aging experiences require moving beyond service-domain frameworks toward spatial stress assessment models. Current research on rural aging in China exhibits methodological gaps, systematically underrepresenting the spatial stressors that older adults face daily, including terrain barriers, infrastructure limitations, climate exposure, and agricultural labor burdens. Existing rural revitalization policies emphasize standardized interventions while inadequately addressing spatial heterogeneity and the spatially-differentiated needs of aging populations. This study developed a GIS-based spatial stress analysis framework that applies Lawton and Nahemow's competence-press model to quantify aging-related stressors and classify rural villages by intervention needs. Using data from 27 villages in Mamuchi Township, Shandong Province, we established four spatial stress indicators: slope gradient index (SGI), solar radiation exposure index (SREI), walkability index (WI), and agricultural intensity index (AII). Analysis of variance and hierarchical clustering revealed significant variation in spatial pressures across villages and identified distinct typologies that require targeted intervention strategies. The framework produces both quantitative stress measurements for individual villages and a classification system that groups villages with similar stress patterns, providing planners and policymakers with practical tools for designing spatially-targeted age-friendly interventions in rural China and similar contexts.

stat.AP↗

arXiv2025

MedPEFT-CL: Dual-Phase Parameter-Efficient Continual Learning with Medical Semantic Adapter and Bidirectional Memory Consolidation

Medical vision-language segmentation models suffer from catastrophic forgetting when adapting to new anatomical structures, requiring complete retraining that limits their clinical deployment. Although continual learning approaches have been studied for various applications, targeted research on continual learning approaches specifically designed for medical vision-language tasks remains underexplored. We propose MedPEFT-CL, a parameter-efficient continual learning framework that addresses both efficient learning of new tasks and preservation of previous knowledge through a dual-phase architecture based on CLIPSeg. Our dual-phase architecture features an adaptive learning phase that employs semantic similarity-based adapter allocation and parameter-efficient fine-tuning for medical tasks through prompt similarity analysis, and a knowledge consolidation phase employing bi-directional Fisher-memory coordination. This creates a reinforcing cycle: consolidation directs replay priorities while new tasks provide challenging samples that improve retention strategies. Our key contributions are: (1) a semantic-driven adapter allocation mechanism that enables efficient learning of new medical tasks, (2) a bi-modal LoRA adaptation that significantly reduces trainable parameters while maintaining cross-modal learning, and (3) bidirectional Fisher-memory coordination that prevents catastrophic forgetting from previous medical tasks. Extensive experiments across diverse medical datasets demonstrate superior forgetting mitigation and performance retention with minimal parameter overhead, making the framework effective for continual learning in medical vision-language scenarios.

cs.CV↗

arXiv2025

SWE-Compass: Towards Unified Evaluation of Agentic Coding Abilities for Large Language Models

Evaluating large language models (LLMs) for software engineering has been limited by narrow task coverage, language bias, and insufficient alignment with real-world developer workflows. Existing benchmarks often focus on algorithmic problems or Python-centric bug fixing, leaving critical dimensions of software engineering underexplored. To address these gaps, we introduce SWE-Compass1, a comprehensive benchmark that unifies heterogeneous code-related evaluations into a structured and production-aligned framework. SWE-Compass spans 8 task types, 8 programming scenarios, and 10 programming languages, with 2000 high-quality instances curated from authentic GitHub pull requests and refined through systematic filtering and validation. We benchmark ten state-of-the-art LLMs under two agentic frameworks, SWE-Agent and Claude Code, revealing a clear hierarchy of difficulty across task types, languages, and scenarios. Moreover, by aligning evaluation with real-world developer practices, SWE-Compass provides a rigorous and reproducible foundation for diagnosing and advancing agentic coding capabilities in large language models.

cs.SE↗

arXiv2025

DeCoRL: Decoupling Reasoning Chains via Parallel Sub-Step Generation and Cascaded Reinforcement for Interpretable and Scalable RLHF

Existing reinforcement learning methods for Chain-of-Thought reasoning suffer from two critical limitations. First, they operate as monolithic black boxes that provide undifferentiated reward signals, obscuring individual step contributions and hindering error diagnosis. Second, sequential decoding has O(n) time complexity. This makes real-time deployment impractical for complex reasoning tasks. We present DeCoRL (Decoupled Reasoning Chains via Coordinated Reinforcement Learning), a novel framework that transforms reasoning from sequential processing into collaborative modular orchestration. DeCoRL trains lightweight specialized models to generate reasoning sub-steps concurrently, eliminating sequential bottlenecks through parallel processing. To enable precise error attribution, the framework designs modular reward functions that score each sub-step independently. Cascaded DRPO optimization then coordinates these rewards while preserving inter-step dependencies. Comprehensive evaluation demonstrates state-of-the-art results across RM-Bench, RMB, and RewardBench, outperforming existing methods including large-scale models. DeCoRL delivers 3.8 times faster inference while maintaining superior solution quality and offers a 22.7\% improvement in interpretability through explicit reward attribution. These advancements, combined with a 72.4\% reduction in energy consumption and a 68\% increase in throughput, make real-time deployment of complex reasoning systems a reality.

cs.CL↗

arXiv2025

MDAS-GNN: Multi-Dimensional Spatiotemporal GNN with Spatial Diffusion for Urban Traffic Risk Forecasting

Traffic accidents represent a critical public health challenge, claiming over 1.35 million lives annually worldwide. Traditional accident prediction models treat road segments independently, failing to capture complex spatial relationships and temporal dependencies in urban transportation networks. This study develops MDAS-GNN, a Multi-Dimensional Attention-based Spatial-diffusion Graph Neural Network integrating three core risk dimensions: traffic safety, infrastructure, and environmental risk. The framework employs feature-specific spatial diffusion mechanisms and multi-head temporal attention to capture dependencies across different time horizons. Evaluated on UK Department for Transport accident data across Central London, South Manchester, and SE Birmingham, MDASGNN achieves superior performance compared to established baseline methods. The model maintains consistently low prediction errors across short, medium, and long-term periods, with particular strength in long-term forecasting. Ablation studies confirm that integrated multi-dimensional features outperform singlefeature approaches, reducing prediction errors by up to 40%. This framework provides civil engineers and urban planners with advanced predictive capabilities for transportation infrastructure design, enabling data-driven decisions for road network optimization, infrastructure resource improvements, and strategic safety interventions in urban development projects.

cs.LG↗

arXiv2025

DPL: Spatial-Conditioned Diffusion Prototype Enhancement for One-Shot Medical Segmentation

One-shot medical image segmentation faces fundamental challenges in prototype representation due to limited annotated data and significant anatomical variability across patients. Traditional prototype-based methods rely on deterministic averaging of support features, creating brittle representations that fail to capture intra-class diversity essential for robust generalization. This work introduces Diffusion Prototype Learning (DPL), a novel framework that reformulates prototype construction through diffusion-based feature space exploration. DPL models one-shot prototypes as learnable probability distributions, enabling controlled generation of diverse yet semantically coherent prototype variants from minimal labeled data. The framework operates through three core innovations: (1) a diffusion-based prototype enhancement module that transforms single support prototypes into diverse variant sets via forward-reverse diffusion processes, (2) a spatial-aware conditioning mechanism that leverages geometric properties derived from prototype feature statistics, and (3) a conservative fusion strategy that preserves prototype fidelity while maximizing representational diversity. DPL ensures training-inference consistency by using the same diffusion enhancement and fusion pipeline in both phases. This process generates enhanced prototypes that serve as the final representations for similarity calculations, while the diffusion process itself acts as a regularizer. Extensive experiments on abdominal MRI and CT datasets demonstrate significant improvements respectively, establishing new state-of-the-art performance in one-shot medical image segmentation.

cs.CV↗

arXiv2025

StructCoh: Structured Contrastive Learning for Context-Aware Text Semantic Matching

Text semantic matching requires nuanced understanding of both structural relationships and fine-grained semantic distinctions. While pre-trained language models excel at capturing token-level interactions, they often overlook hierarchical structural patterns and struggle with subtle semantic discrimination. In this paper, we proposed StructCoh, a graph-enhanced contrastive learning framework that synergistically combines structural reasoning with representation space optimization. Our approach features two key innovations: (1) A dual-graph encoder constructs semantic graphs via dependency parsing and topic modeling, then employs graph isomorphism networks to propagate structural features across syntactic dependencies and cross-document concept nodes. (2) A hierarchical contrastive objective enforces consistency at multiple granularities: node-level contrastive regularization preserves core semantic units, while graph-aware contrastive learning aligns inter-document structural semantics through both explicit and implicit negative sampling strategies. Experiments on three legal document matching benchmarks and academic plagiarism detection datasets demonstrate significant improvements over state-of-the-art methods. Notably, StructCoh achieves 86.7% F1-score (+6.2% absolute gain) on legal statute matching by effectively identifying argument structure similarities.

cs.CL↗

arXiv2025

Quasi-Isometric Reductions Between Infinite Strings

This paper studies the recursion-theoretic aspects of large-scale geometries of infinite strings, a subject initiated by Khoussainov and Takisaka (2017). We investigate several notions of quasi-isometric reductions between recursive infinite strings and prove various results on the equivalence classes of such reductions. The main result is the construction of two infinite recursive strings $α$ and $β$ such that $α$ is strictly quasi-isometrically reducible to $β$, but the reduction cannot be made recursive. This answers an open problem posed by Khoussainov and Takisaka.

cs.FL↗

arXiv2025

DDiT: Dynamic Resource Allocation for Diffusion Transformer Model Serving

The Text-to-Video (T2V) model aims to generate dynamic and expressive videos from textual prompts. The generation pipeline typically involves multiple modules, such as language encoder, Diffusion Transformer (DiT), and Variational Autoencoders (VAE). Existing serving systems often rely on monolithic model deployment, while overlooking the distinct characteristics of each module, leading to inefficient GPU utilization. In addition, DiT exhibits varying performance gains across different resolutions and degrees of parallelism, and significant optimization potential remains unexplored. To address these problems, we present DDiT, a flexible system that integrates both inter-phase and intra-phase optimizations. DDiT focuses on two key metrics: optimal degree of parallelism, which prevents excessive parallelism for specific resolutions, and starvation time, which quantifies the sacrifice of each request. To this end, DDiT introduces a decoupled control mechanism to minimize the computational inefficiency caused by imbalances in the degree of parallelism between the DiT and VAE phases. It also designs a greedy resource allocation algorithm with a novel scheduling mechanism that operates at the single-step granularity, enabling dynamic and timely resource scaling. Our evaluation on the T5 encoder, OpenSora SDDiT, and OpenSora VAE models across diverse datasets reveals that DDiT significantly outperforms state-of-the-art baselines by up to 1.44x in p99 latency and 1.43x in average latency.

cs.DC↗

arXiv2022

Alternating Automatic Register Machines

This paper introduces and studies a new model of computation called an Alternating Automatic Register Machine (AARM). An AARM possesses the basic features of a conventional register machine and an alternating Turing machine, but can carry out computations using bounded automatic relations in a single step. One finding is that an AARM can recognise some NP-complete problems, including SAT (using a particular coding), in $\log^* n + O(1)$ steps. On the other hand, if all problems in P can be solved by an AARM in $O(\log^*n)$ rounds, then $\text{P} \subset \text{PSPACE}$. Furthermore, we study an even more computationally powerful machine, called a Polynomial-Size Padded Alternating Automatic Register Machine (PAARM), which allows the input to be padded with a polynomial-size string. It is shown that the polynomial hierarchy can be characterised as the languages that are recognised by a PAARM in $\log^*n + O(1)$ steps. These results illustrate the power of alternation when combined with computations involving automatic relations, and uncover a finer gradation between known complexity classes.

cs.CC↗

arXiv2022

A Computation Model with Automatic Functions and Relations as Primitive Operations

Prior work of Hartmanis and Simon (Hartmanis and Simon, 1974) and Floyd and Knuth (Floyd and Knuth, 1990) investigated what happens if a device uses primitive steps more natural than single updates of a Turing tape. One finding was that in the numerical setting, addition, subtraction, comparisons and bit-wise Boolean operations of numbers preserve polynomial time while incorporating concatenation or multiplication allows to solve all PSPACE problems in polynomially many steps. Therefore we propose to use updates and comparisons with automatic functions as primitive operations and use constantly many registers; the resulting model covers all primitive operations of Hartmanis and Simon as well as Floyd and Knuth, but the model remains in polynomial time. The present work investigates in particular the deterministic complexity of various natural problems and also gives an overview on the nondeterministic complexity of this model.

cs.CC↗

arXiv2021

Learnability and Positive Equivalence Relations

Prior work of Gavryushkin, Khoussainov, Jain and Stephan investigated what algebraic structures can be realised in worlds given by a positive (= recursively enumerable) equivalence relation which partitions the natural numbers into infinitely many equivalence classes. The present work investigates the infinite one-one numbered recursively enumerable (r.e.) families realised by such relations and asks how the choice of the equivalence relation impacts the learnability properties of these classes when studying learnability in the limit from positive examples, also known as learning from text. For all choices of such positive equivalence relations, for each of the following entries, there are one-one numbered r.e. families which satisfy it: (a) they are behaviourally correctly learnable but not vacillatorily learnable; (b) they are explanatorily learnable but not confidently learnable; (c) they are not behaviourally correctly learnable. Furthermore, there is a positive equivalence relation which enforces that (d) every vacillatorily learnable one-one numbered family of languages closed under this equivalence relation is already explanatorily learnable and cannot be confidently learnable.

cs.LO↗

arXiv2021

Ordered Semiautomatic Rings with Applications to Geometry

The present work looks at semiautomatic rings with automatic addition and comparisons which are dense subrings of the real numbers and asks how these can be used to represent geometric objects such that certain operations and transformations are automatic. The underlying ring has always to be a countable dense subring of the real numbers and additions and comparisons and multiplications with constants need to be automatic. It is shown that the ring can be selected such that equilateral triangles can be represented and rotations by 30 degrees are possible, while the standard representation of the b-adic rationals does not allow this.

cs.FL↗

arXiv2019

The teaching complexity of erasing pattern languages with bounded variable frequency

Patterns provide a concise, syntactic way of describing a set of strings, but their expressive power comes at a price: a number of fundamental decision problems concerning (erasing) pattern languages, such as the membership problem and inclusion problem, are known to be NP-complete or even undecidable, while the decidability of the equivalence problem is still open; in learning theory, the class of pattern languages is unlearnable in models such as the distribution-free (PAC) framework (if $\mathcal{P}/poly \neq \mathcal{NP}/poly$). Much work on the algorithmic learning of pattern languages has thus focussed on interesting subclasses of patterns for which positive learnability results may be achieved. A natural restriction on a pattern is a bound on its variable frequency -- the maximum number $m$ such that some variable occurs exactly $m$ times in the pattern. This paper examines the effect of limiting the variable frequency of all patterns belonging to a class $Π$ on the worst-case minimum number of labelled examples needed to uniquely identify any pattern of $Π$ in cooperative teaching-learning models. Two such models, the teaching dimension model as well as the preference-based teaching model, will be considered.

cs.FL↗

arXiv2019

Random Subgroups of Rationals

This paper introduces and studies a notion of \emph{algorithmic randomness} for subgroups of rationals. Given a randomly generated additive subgroup $(G,+)$ of rationals, two main questions are addressed: first, what are the model-theoretic and recursion-theoretic properties of $(G,+)$; second, what learnability properties can one extract from $G$ and its subclass of finitely generated subgroups? For the first question, it is shown that the theory of $(G,+)$ coincides with that of the additive group of integers and is therefore decidable; furthermore, while the word problem for $G$ with respect to any generating sequence for $G$ is not even semi-decidable, one can build a generating sequence $β$ such that the word problem for $G$ with respect to $β$ is co-recursively enumerable (assuming that the set of generators of $G$ is limit-recursive). In regard to the second question, it is proven that there is a generating sequence $β$ for $G$ such that every non-trivial finitely generated subgroup of $G$ is recursively enumerable and the class of all such subgroups of $G$ is behaviourally correctly learnable, that is, every non-trivial finitely generated subgroup can be semantically identified in the limit (again assuming that the set of generators of $G$ is limit-recursive). On the other hand, the class of non-trivial finitely generated subgroups of $G$ cannot be syntactically identified in the limit with respect to any generating sequence for $G$. The present work thus contributes to a recent line of research studying algorithmically random infinite structures and uncovers an interesting connection between the arithmetical complexity of the set of generators of a randomly generated subgroup of rationals and the learnability of its finitely generated subgroups.

cs.LO↗