Author name cluster

Xinyuan Wang

Possible papers associated with this exact author name in Arrow. This page groups case-insensitive exact name matches and is not a full identity disambiguation profile.

14 papers

2 author rows

AAAI Conference 2026 Conference Paper

Brownian Bridge Augmented Surrogate Simulation and Injection Planning for Geological CO2 Storage

Haoyue Bai
Guodong Chen
Wangyang Ying
Xinyuan Wang
Nanxu Gong
Sixun Dong
Giulia Pedrielli
Haoyu Wang

Geological CO2 storage (GCS) involves injecting captured CO2 into deep subsurface formations to support climate goals. The effective management of GCS relies on adaptive injection planning to dynamically control injection rates and well pressures to balance both storage safety and efficiency. Prior literature, including numerical optimization methods and surrogate-optimization methods, is limited by real-world GCS requirements of smooth state transitions and goal-directed planning within limited time. To address these limitations, we propose a Brownian Bridge–augmented framework for surrogate simulation and injection planning in GCS and develop two insights (i) Brownian bridge as smooth state regularizer for better surrogate simulator; (ii) Brownian bridge as goal-time-conditioned planning guidance for better injection planning. Our method has three stages: (i) learning deep Brownian bridge representations with contrastive and reconstructive losses from historical reservoir and utility trajectories, (ii) incorporating Brownian bridge-based next state interpolation for simulator regularization (iii) guiding injection planning with Brownian utility-conditioned trajectories to generate high-quality injection plans. Experimental results across multiple datasets collected from diverse GCS settings demonstrate that our framework consistently improves simulation fidelity and planning effectiveness while maintaining low computational overhead.

PDF Details DOI

AAAI Conference 2026 Conference Paper

Efficient Post-Training Refinement of Latent Reasoning in Large Language Models

Xinyuan Wang
Dongjie Wang
Wangyang Ying
Haoyue Bai
Nanxu Gong
Sixun Dong
Kunpeng Liu
Yanjie Fu

Reasoning is a key component of language understanding in Large Language Models. While Chain-of-Thought prompting enhances performance via explicit intermediate steps, it suffers from sufficient token overhead and a fixed reasoning trajectory, preventing step-wise refinement. Recent advances in latent reasoning address these limitations by refining internal reasoning processes directly in the model’s latent space, without producing explicit outputs. However, a key challenge remains: how to effectively update reasoning embeddings during post-training to guide the model toward more accurate solutions. To overcome this challenge, we propose a lightweight post-training framework that refines latent reasoning trajectories using two novel strategies: (1) Contrastive reasoning feedback, which compares reasoning embeddings against strong and weak baselines to infer effective update directions via embedding enhancement; (2) Residual embedding refinement, which stabilizes updates by progressively integrating current and historical gradients, enabling fast yet controlled convergence. Extensive experiments and case studies are conducted on five reasoning benchmarks to demonstrate the effectiveness of the proposed framework. Notably, a +5% accuracy gain on MathQA without additional training.

PDF Details DOI

EAAI Journal 2026 Journal Article

Enhancing remote sensing object detection via selective-perspective-class integration

Xinyuan Wang
Lian Peng
Xiangcheng Li
Yilin He
KinTak U

Object detection in remote sensing imagery is persistently challenged by extreme scale variations, densely distributed objects, and cluttered backgrounds. Although modern detectors like YOLOv8 have shown promising results, their backbone networks often lack explicit mechanisms to guide multi-scale feature refinement, which constrains their performance on high-resolution aerial imagery. In this work, we propose You Only Look Once-Selective-Perspective-Class Integration (YOLO-SPCI), an attention-enhanced detection framework that introduces a lightweight Selective-Perspective-Class Integration (SPCI) module to improve feature representation. The SPCI module integrates three components: a Selective Stream Gate (SSG) for adaptive regulation of global feature flow, a Perspective Fusion Module (PFM) for context-aware multi-scale integration, and a Class Discrimination Module (CDM) to enhance inter-class separability. We embed two SPCI blocks into the P3 and P5 stages of the YOLOv8 backbone, facilitating effective multi-scale feature refinement while maintaining full compatibility with the original neck and head structures. Experiments on the Northwestern Polytechnical University Very High Resolution-10 dataset (NWPU VHR-10), Dataset for Object Detection in Optical Remote Sensing Images (DIOR), and Remote Sensing Object Detection dataset (RSOD) demonstrate that YOLO-SPCI achieves superior performance compared to state-of-the-art detectors.

Details DOI

EAAI Journal 2026 Journal Article

Zero-velocity update -aided navigation method for miniature quadruped robot based on adapted virtual inertial measurement unit

Siwei Tang
Weixing Qian
Sen Wang
Feng Yang
Xinyuan Wang
Weinan Gao
Pengyu Liu

Addressing the challenges associated with installing inertial measurement units (IMUs) on the feet of miniature quadruped robots, this paper proposes a zero-velocity update (ZUPT) method based on adaptive virtual inertial measurement unit (VIMU). This approach eliminates the reliance of existing ZUPT method for inertial navigation systems on foot-mounted IMUs and gait recognition. By utilizing the IMU outputs from legs and feet of a quadruped robot as the training dataset, an innovative Convolutional Neural Network (CNN)- Bidirectional gated recurrent unit neural network (BiGRU)-Attention hybrid network is constructed to establish a nonlinear mapping relationship between the multiple IMUs. In practical applications, the foot-mounted VIMU can be generated solely from the leg-mounted IMU data, and the modified navigation parameters are then output through a ZUPT algorithm to achieve accurate positioning of the quadruped robot. Experimental results demonstrate that the positioning errors of this method is about 1. 34 % of the total path under diverse terrain conditions, including slopes, stairs, and grasslands, outperforming gait recognition-dependent methods in terms of accuracy. This approach effectively implements the inertial navigation function of quadruped robots and enhances the adaptability of ZUPT method to unstructured and unknown terrains. This approach has great potential to improve Global Navigation Satellite Systems (GNSS)-denied positioning performance of quadruped robots in complex environments without the assistance of visual sensor and LightLaser Detection and Ranging (LiDAR).

Details DOI

ICML Conference 2025 Conference Paper

We propose Network Automatic Relevance Determination (NARD), an extension of ARD for linearly probabilistic models, to simultaneously model sparse relationships between inputs $X \in \mathbb R^{d \times N}$ and outputs $Y \in \mathbb R^{m \times N}$, while capturing the correlation structure among the $Y$. NARD employs a matrix normal prior which contains a sparsity-inducing parameter to identify and discard irrelevant features, thereby promoting sparsity in the model. Algorithmically, it iteratively updates both the precision matrix and the relationship between $Y$ and the refined inputs. To mitigate the computational inefficiencies of the $\mathcal O(m^3 + d^3)$ cost per iteration, we introduce Sequential NARD, which evaluates features sequentially, and a Surrogate Function Method, leveraging an efficient approximation of the marginal likelihood and simplifying the calculation of determinant and inverse of an intermediate matrix. Combining the Sequential update with the Surrogate Function method further reduces computational costs. The computational complexity per iteration for these three methods is reduced to $\mathcal O(m^3+p^3)$, $\mathcal O(m^3 + d^2)$, $\mathcal O(m^3+p^2)$ respectively, where $p \ll d$ is the final number of features in the model. Our methods demonstrate significant improvements in computational efficiency with comparable performance on both synthetic and real-world datasets.

Details

ICLR Conference 2025 Conference Paper

LeFusion: Controllable Pathology Synthesis via Lesion-Focused Diffusion Models

Hantao Zhang
Yuhe Liu
Jiancheng Yang
Shouhong Wan
Xinyuan Wang
Wei Peng
Pascal Fua

Patient data from real-world clinical practice often suffers from data scarcity and long-tail imbalances, leading to biased outcomes or algorithmic unfairness. This study addresses these challenges by generating lesion-containing image-segmentation pairs from lesion-free images. Previous efforts in medical imaging synthesis have struggled with separating lesion information from background, resulting in low-quality backgrounds and limited control over the synthetic output. Inspired by diffusion-based image inpainting, we propose LeFusion, a lesion-focused diffusion model. By redesigning the diffusion learning objectives to focus on lesion areas, we simplify the learning process and improve control over the output while preserving high-fidelity backgrounds by integrating forward-diffused background contexts into the reverse diffusion process. Additionally, we tackle two major challenges in lesion texture synthesis: 1) multi-peak and 2) multi-class lesions. We introduce two effective strategies: histogram-based texture control and multi-channel decomposition, enabling the controlled generation of high-quality lesions in difficult scenarios. Furthermore, we incorporate lesion mask diffusion, allowing control over lesion size, location, and boundary, thus increasing lesion diversity. Validated on 3D cardiac lesion MRI and lung nodule CT datasets, LeFusion-generated data significantly improves the performance of state-of-the-art segmentation models, including nnUNet and SwinUNETR.

Details

TIST Journal 2025 Journal Article

LLM-Enhanced User–Item Interactions: Leveraging Edge Information for Optimized Recommendations

Xinyuan Wang
Liang Wu
Liangjie Hong
Hao Liu
Yanjie Fu

Graph recommendation methods, representing a connected interaction perspective, reformulate user–item interactions as graphs to leverage graph structure and topology to recommend and have proved practical effectiveness at scale. Large language models (LLMs), representing a textual generative perspective, excel at modeling user languages, understanding behavioral contexts, capturing user–item semantic relationships, analyzing textual sentiments, and generating coherent and contextually relevant texts as recommendations. However, there is a gap between the connected graph perspective and the text generation perspective as the task formulations are different. A research question arises: how can we effectively integrate the two perspectives for more personalized RecSys? To fill this gap, we propose to incorporate graph-edge information into LLMs via prompt and attention innovations. We reformulate recommendations as a probabilistic generative problem using prompts. We develop a framework to incorporate graph edge information from the prompt and attention mechanisms for graph-structured LLM recommendations. We develop a new prompt design that brings in both first-order and second-order graph relationships; we devise an improved LLM attention mechanism to embed direct the spatial and connectivity information of edges. Our evaluation of real-world datasets demonstrates the framework’s ability to understand connectivity information in graph data and to improve the relevance and quality of recommendation results. Our code is released at: https://github.com/anord-wang/LLM4REC.git.

Details DOI

AAAI Conference 2025 Conference Paper

Medical MLLM Is Vulnerable: Cross-Modality Jailbreak and Mismatched Attacks on Medical Multimodal Large Language Models

Xijie Huang
Xinyuan Wang
Hantao Zhang
Yinghao Zhu
Jiawen Xi
Jingkun An
Hao Wang
Hao Liang

Security concerns related to Large Language Models (LLMs) have been extensively explored; however, the safety implications for Multimodal Large Language Models (MLLMs), particularly in medical contexts (MedMLLMs), remain inadequately addressed. This paper investigates the security vulnerabilities of MedMLLMs, focusing on their deployment in clinical environments where the accuracy and relevance of question-and-answer interactions are crucial for addressing complex medical challenges. We introduce and redefine two attack types: mismatched malicious attack (2M-attack) and optimized mismatched malicious attack (O2M-attack), by integrating existing clinical data with atypical natural phenomena. Using the comprehensive 3MAD dataset that we developed, which spans a diverse range of medical imaging modalities and adverse medical scenarios, we performed an in-depth analysis and proposed the MCM optimization method. This approach significantly improves the attack success rate against MedMLLMs. Our evaluations, which include white-box attacks on LLaVA-Med and transfer (black-box) attacks on four other SOTA models, reveal that even MedMLLMs designed with advanced security mechanisms remain vulnerable to breaches. This study highlights the critical need for robust security measures to enhance the safety and reliability of open-source MedMLLMs, especially in light of the potential impact of jailbreak attacks and other malicious exploits in clinical applications. Warning: Medical jailbreaking may generate content that includes unverified diagnoses and treatment recommendations. Always consult professional medical advice.

PDF Details DOI

NeurIPS Conference 2025 Conference Paper

OpenCUA: Open Foundations for Computer-Use Agents

Xinyuan Wang
Bowen Wang
Dunjie Lu
Junlin Yang
Tianbao Xie
Junli Wang
Jiaqi Deng
Xiaole Guo

Vision-language models have demonstrated impressive capabilities as computer-use agents (CUAs) capable of automating diverse computer tasks. As their commercial potential grows, critical details of the most capable CUA systems remain closed. As these agents will increasingly mediate digital interactions and execute consequential decisions on our behalf, the research community needs access to open CUA frameworks to study their capabilities, limitations, and risks. To bridge this gap, we propose OpenCUA, a comprehensive open-source framework for scaling CUA data and foundation models. Our framework consists of: (1) an annotation infrastructure that seamlessly captures human computer-use demonstrations; (2) AgentNet, the first large-scale computer-use task dataset spanning 3 operating systems and 200+ applications and websites; (3) a scalable pipeline that transforms demonstrations into state–action pairs with reflective long Chain-of-Thought reasoning that sustain robust performance gains as data scales. Our end-to-end agent models demonstrate strong performance across CUA benchmarks. In particular, OpenCUA-72B achieves an average success rate of 45. 0% on OSWorld‑Verified, establishing a new state-of-the-art (SOTA) among open-source models. Further analysis confirms that our approach generalizes well across domains and benefits significantly from increased test-time computation. We release our annotation tool, datasets, code, and models to build open foundations for further CUA research.

PDF Details

NeurIPS Conference 2025 Conference Paper

Scaling Computer-Use Grounding via User Interface Decomposition and Synthesis

Tianbao Xie
Jiaqi Deng
Xiaochuan Li
Junlin Yang
Haoyuan Wu
Jixuan Chen
Wenjing Hu
Xinyuan Wang

Graphical user interface (GUI) grounding, the ability to map natural language instructions to specific actions on graphical user interfaces, remains a critical bottleneck in computer use agent development. Current benchmarks oversimplify grounding tasks as short referring expressions, failing to capture the complexity of real-world interactions that require software commonsense, layout understanding, and fine-grained manipulation capabilities. To address these limitations, we introduce OSWorld-G, a comprehensive benchmark comprising 564 finely annotated samples across diverse task types including text matching, element recognition, layout understanding, and precise manipulation. Additionally, we synthesize and release the largest computer use grounding dataset Jedi, which contains 4 million examples through multi-perspective decoupling of tasks. Our multi-scale models trained on Jedi demonstrate its effectiveness by outperforming existing approaches on ScreenSpot-v2, ScreenSpot-Pro, and our OSWorld-G. Furthermore, we demonstrate that improved grounding with Jedi directly enhances agentic capabilities of general foundation models on complex computer tasks with state-of-the-art performance, improving from 23% to 51% on OSWorld. Through detailed ablation studies, we identify key factors contributing to grounding performance and verify that combining specialized data for different interface elements enables compositional generalization to novel interfaces. All benchmark, data, checkpoints, and code are open-sourced and available at https: //osworld-grounding. github. io.

PDF Details

NeurIPS Conference 2025 Conference Paper

Sculpting Features from Noise: Reward-Guided Hierarchical Diffusion for Task-Optimal Feature Transformation

Nanxu Gong
Zijun Li
Sixun Dong
Haoyue Bai
Wangyang Ying
Xinyuan Wang
Yanjie Fu

Feature Transformation (FT) crafts new features from original ones via mathematical operations to enhance dataset expressiveness for downstream models. However, existing FT methods exhibit critical limitations: discrete search struggles with enormous combinatorial spaces, impeding practical use; and continuous search, being highly sensitive to initialization and step sizes, often becomes trapped in local optima, restricting global exploration. To overcome these limitations, DIFFT redefines FT as a reward-guided generative task. It first learns a compact and expressive latent space for feature sets using a Variational Auto-Encoder (VAE). A Latent Diffusion Model (LDM) then navigates this space to generate high-quality feature embeddings, its trajectory guided by a performance evaluator towards task-specific optima. This synthesis of global distribution learning (from LDM) and targeted optimization (reward guidance) produces potent embeddings, which a novel semi-autoregressive decoder efficiently converts into structured, discrete features, preserving intra-feature dependencies while allowing parallel inter-feature generation. Extensive experiments on 14 benchmark datasets show DIFFT consistently outperforms state-of-the-art baselines in predictive accuracy and robustness, with significantly lower training and inference times.

PDF Details

IJCAI Conference 2025 Conference Paper

Unsupervised Feature Transformation via In-context Generation, Generator-critic LLM Agents, and Duet-play Teaming

Nanxu Gong
Xinyuan Wang
Wangyang Ying
Haoyue Bai
Sixun Dong
Haifeng Chen
Yanjie Fu

Feature transformation involves generating a new set of features from the original dataset to enhance the data's utility. In certain domains like material performance screening, dimensionality is large and collecting labels is expensive and lengthy. It highly necessitates transforming feature spaces efficiently and without supervision to enhance data readiness and AI utility. However, existing methods fall short in efficient navigation of a vast space of feature combinations, and are mostly designed for supervised settings. To fill this gap, our unique perspective is to leverage a generator-critic duet-play teaming framework using LLM agents and in-context learning to derive pseudo-supervision from unsupervised data. The framework consists of three interconnected steps: (1) Critic agent diagnoses data to generate actionable advice, (2) Generator agent produces tokenized feature transformations guided by the critic's advice, and (3) Iterative refinement ensures continuous improvement through feedback between agents. The generator-critic framework can be generalized to human-agent collaborative generation, by replacing the critic agent with human experts. Extensive experiments demonstrate that the proposed framework outperforms even supervised baselines in feature transformation efficiency, robustness, and practical applicability across diverse datasets. Our code is publicly available at https: //github. com/NanxuGong/LPFG.

PDF Details DOI

AAAI Conference 2024 Conference Paper

BiPFT: Binary Pre-trained Foundation Transformer with Low-Rank Estimation of Binarization Residual Polynomials

Xingrun Xing
Li Du
Xinyuan Wang
Xianlin Zeng
Yequan Wang
Zheng Zhang
Jiajun Zhang

Pretrained foundation models offer substantial benefits for a wide range of downstream tasks, which can be one of the most potential techniques to access artificial general intelligence. However, scaling up foundation transformers for maximal task-agnostic knowledge has brought about computational challenges, especially on resource-limited devices such as mobiles. This work proposes the first Binary Pretrained Foundation Transformer (BiPFT) for natural language understanding (NLU) tasks, which remarkably saves 56 times operations and 28 times memory. In contrast to previous task-specific binary transformers, BiPFT exhibits a substantial enhancement in the learning capabilities of binary neural networks (BNNs), promoting BNNs into the era of pre-training. Benefiting from extensive pretraining data, we further propose a data-driven binarization method. Specifically, we first analyze the binarization error in self-attention operations and derive the polynomials of binarization error. To simulate full-precision self-attention, we define binarization error as binarization residual polynomials, and then introduce low-rank estimators to model these polynomials. Extensive experiments validate the effectiveness of BiPFTs, surpassing task-specific baseline by 15.4% average performance on the GLUE benchmark. BiPFT also demonstrates improved robustness to hyperparameter changes, improved optimization efficiency, and reduced reliance on downstream distillation, which consequently generalize on various NLU tasks and simplify the downstream pipeline of BNNs. Our code and pretrained models are publicly available at https://github.com/Xingrun-Xing/BiPFT.

PDF Details DOI

TAAS Journal 2014 Journal Article

A Host-Based Approach for Unknown Fast-Spreading Worm Detection and Containment

Songqing Chen
Lei Liu
Xinyuan Wang
Xinwen Zhang
Zhao Zhang

The fast-spreading worm, which immediately propagates itself after a successful infection, is becoming one of the most serious threats to today’s networked information systems. In this article, we present WormTerminator, a host-based solution for fast Internet worm detection and containment with the assistance of virtual machine techniques based on the fast-worm defining characteristic. In WormTerminator, a virtual machine cloning the host OS runs in parallel to the host OS. Thus, the virtual machine has the same set of vulnerabilities as the host. Any outgoing traffic from the host is diverted through the virtual machine. If the outgoing traffic from the host is for fast worm propagation, the virtual machine should be infected and will exhibit worm propagation pattern very quickly because a fast-spreading worm will start to propagate as soon as it successfully infects a host. To prove the concept, we have implemented a prototype of WormTerminator and have examined its effectiveness against the real Internet worm Linux/Slapper. Our empirical results confirm that WormTerminator is able to completely contain worm propagation in real-time without blocking any non-worm traffic. The major performance cost of WormTerminator is a one-time delay to the start of each outgoing normal connection for worm detection. To reduce the performance overhead, caching is utilized, through which WormTerminator will delay no more than 6% normal outgoing traffic for such detection on average.

Details DOI