Xiaowen Zhang

Xuanlong Yu, Yi Zuo, Zitao Wang et al. · cmu

This paper outlines the winning solutions employed in addressing the MUAD uncertainty quantification challenge held at ICCV 2023. The challenge was centered around semantic segmentation in urban environments, with a particular focus on natural adversarial scenarios. The report presents the results of 19 submitted entries, with numerous techniques drawing inspiration from cutting-edge uncertainty quantification methodologies presented at prominent conferences in the fields of computer vision and machine learning and journals over the past few years. Within this document, the challenge is introduced, shedding light on its purpose and objectives, which primarily revolved around enhancing the robustness of semantic segmentation in urban scenes under varying natural adversarial conditions. The report then delves into the top-performing solutions. Moreover, the document aims to provide a comprehensive overview of the diverse solutions deployed by all participants. By doing so, it seeks to offer readers a deeper insight into the array of strategies that can be leveraged to effectively handle the inherent uncertainties associated with autonomous driving and semantic segmentation, especially within urban environments.

Mengli Cheng, Yue Gao, Guoqiang Liu et al.

We present EasyRec, an easy-to-use, extendable and efficient recommendation framework for building industrial recommendation systems. Our EasyRec framework is superior in the following aspects: first, EasyRec adopts a modular and pluggable design pattern to reduce the efforts to build custom models; second, EasyRec implements hyper-parameter optimization and feature selection algorithms to improve model performance automatically; third, EasyRec applies online learning to fast adapt to the ever-changing data distribution. The code is released: https://github.com/alibaba/EasyRec.

Diana M. Negoescu, Humberto Gonzalez, Saad Eddin Al Orjany et al. · cmu

We introduce Epsilon*, a new privacy metric for measuring the privacy risk of a single model instance prior to, during, or after deployment of privacy mitigation strategies. The metric requires only black-box access to model predictions, does not require training data re-sampling or model re-training, and can be used to measure the privacy risk of models not trained with differential privacy. Epsilon* is a function of true positive and false positive rates in a hypothesis test used by an adversary in a membership inference attack. We distinguish between quantifying the privacy loss of a trained model instance, which we refer to as empirical privacy, and quantifying the privacy loss of the training mechanism which produces this model instance. Existing approaches in the privacy auditing literature provide lower bounds for the latter, while our metric provides an empirical lower bound for the former by relying on an ($ε$, $δ$)-type of quantification of the privacy of the trained model instance. We establish a relationship between these lower bounds and show how to implement Epsilon* to avoid numerical and noise amplification instability. We further show in experiments on benchmark public data sets that Epsilon* is sensitive to privacy risk mitigation by training with differential privacy (DP), where the value of Epsilon* is reduced by up to 800% compared to the Epsilon* values of non-DP trained baseline models. This metric allows privacy auditors to be independent of model owners, and enables visualizing the privacy-utility landscape to make informed decisions regarding the trade-offs between model privacy and utility.

Furkan Kınlı, Sami Menteş, Barış Özcan et al.

This paper introduces the methods and the results of AIM 2022 challenge on Instagram Filter Removal. Social media filters transform the images by consecutive non-linear operations, and the feature maps of the original content may be interpolated into a different domain. This reduces the overall performance of the recent deep learning strategies. The main goal of this challenge is to produce realistic and visually plausible images where the impact of the filters applied is mitigated while preserving the content. The proposed solutions are ranked in terms of the PSNR value with respect to the original images. There are two prior studies on this task as the baseline, and a total of 9 teams have competed in the final phase of the challenge. The comparison of qualitative results of the proposed solutions and the benchmark for the challenge are presented in this report.

Dihan Zheng, Xiaowen Zhang, Kaisheng Ma et al.

Collecting paired training data is difficult in practice, but the unpaired samples broadly exist. Current approaches aim at generating synthesized training data from unpaired samples by exploring the relationship between the corrupted and clean data. This work proposes LUD-VAE, a deep generative method to learn the joint probability density function from data sampled from marginal distributions. Our approach is based on a carefully designed probabilistic graphical model in which the clean and corrupted data domains are conditionally independent. Using variational inference, we maximize the evidence lower bound (ELBO) to estimate the joint probability density function. Furthermore, we show that the ELBO is computable without paired samples under the inference invariant assumption. This property provides the mathematical rationale of our approach in the unpaired setting. Finally, we apply our method to real-world image denoising, super-resolution, and low-light image enhancement tasks and train the models using the synthetic data generated by the LUD-VAE. Experimental results validate the advantages of our method over other approaches.

Xiaowen Zhang, Zijie Yue, Yong Luo et al.

Object counting is a fundamental task in computer vision, with broad applicability in many real-world scenarios. Fully-supervised counting methods require costly point-level annotations per object. Few weakly-supervised methods leverage only image-level object counts as supervision and achieve fairly promising results. They are, however, often limited to counting a single category, e.g. person. In this paper, we propose WS-COC, the first MLLM-driven weakly-supervised framework for class-agnostic object counting. Instead of directly fine-tuning MLLMs to predict object counts, which can be challenging due to the modality gap, we incorporate three simple yet effective strategies to bootstrap the counting paradigm in both training and testing: First, a divide-and-discern dialogue tuning strategy is proposed to guide the MLLM to determine whether the object count falls within a specific range and progressively break down the range through multi-round dialogue. Second, a compare-and-rank count optimization strategy is introduced to train the MLLM to optimize the relative ranking of multiple images according to their object counts. Third, a global-and-local counting enhancement strategy aggregates and fuses local and global count predictions to improve counting performance in dense scenes. Extensive experiments on FSC-147, CARPK, PUCPR+, and ShanghaiTech show that WS-COC matches or even surpasses many state-of-art fully-supervised methods while significantly reducing annotation costs. Code is available at https://github.com/viscom-tongji/WS-COC.

Xiaowen Zhang, Zhi Gao, Licheng Jiao et al.

In vision-language models (VLMs), misalignment between textual descriptions and visual coordinates often induces hallucinations. This issue becomes particularly severe in dense prediction tasks such as spatial-temporal video grounding (STVG). Prior approaches typically focus on enhancing visual-textual alignment or attaching auxiliary decoders. However, these strategies inevitably introduce additional trainable modules, leading to significant annotation costs and computational overhead. In this work, we propose a novel visual prompting paradigm that avoids the difficult problem of aligning coordinates across modalities. Specifically, we reformulate per-frame coordinate prediction as a compact instance-level identification problem by assigning each object a unique, temporally consistent ID. These IDs are embedded into the video as visual prompts, providing explicit and interpretable inputs to the VLMs. Furthermore, we introduce STVG-R1, the first reinforcement learning framework for STVG, which employs a task-driven reward to jointly optimize temporal accuracy, spatial consistency, and structural format regularization. Extensive experiments on six benchmarks demonstrate the effectiveness of our approach. STVG-R1 surpasses the baseline Qwen2.5-VL-7B by a remarkable margin of 20.9% on m_IoU on the HCSTVG-v2 benchmark, establishing a new state of the art (SOTA). Surprisingly, STVG-R1 also exhibits strong zero-shot generalization to multi-object referring video object segmentation tasks, achieving a SOTA 47.3% J&F on MeViS.

Chi Sing Chum, Jia Lu, Claire Tang et al.

We design and develop a secret-sharing-scheme-based cyberattack detection model(S3CDM)that can detect unauthorized or illegal activities (especially insider attacks) and protect sensitive information within complex network infrastructures of large organizations. The model splits a secret among a group of legitimate participants or components for authentication, integration and detection of unauthorized activities. Traditional Shamir's polynomial interpolation based and our own hash function based schemes are utilized in the model, they both are practical and efficient to make sure the communications between different components are secure and any unauthorized activities can be detected. The model offers a flexible multi-factor authentication method to enhance the overall system security. Probability analysis [3] shows that multiple component model is more resistant against cyberattacks than the single component one. To demonstrate the feasibility, we implement the S3CDM in three parts on Google Cloud Platform, i.e., the front end UI (User Interface) running on an HTTP server, the back end individual services written in Python, and a PostgreSQL database. Docker is used to manage the start and stop of individual services and their URLs. We demonstrate how to use the UI with a use case of simulation of broken path in details.

Xiaowen Zhang, Huaien Zhang, Shin Hwei Tan

Software redesign preserves functionality while improving quality attributes, but manual reuse of code and tests is costly and error-prone, especially in crossrepository redesigns. Focusing on static analyzers where cross-repo redesign needs often arise, we conduct a bidirectional study of the ongoing Soot/SootUp redesign case using an action research methodology that combines empirical investigation with validated open-source contributions. Our study reveals: (1) non-linear migration which necessitates bidirectional reuse, (2) deferred reuse via TODOs, (3) neglected test porting, and (4) residual bug propagation during migrations. We identify tracking corresponding code and tests as the key challenge, and address it by retrofitting clone detection to derive code mappings between original and redesigned projects. Guided by semantic reuse patterns derived in our study, we propose Semantic Alignment Heuristics and a scalable hierarchical detection strategy. Evaluations on two redesigned project pairs (Soot/SootUp and FindBugs/SpotBugs) show that our approach achieves an average reduction of 33-99% in likely irrelevant clones at a SAS threshold of 0.5 across all tool results, and improves precision up to 86% on our benchmark of 1,749 samples. Moreover, we contribute to the redesigned projects by submitting five issues and 10 pull requests, of which eight have been merged.

Xiaowen Zhang, Zhenyu Bi, Patrick Lachance et al.

As cosmological simulations and their associated software become increasingly complex, physicists face the challenge of searching through vast amounts of literature and user manuals to extract simulation parameters from dense academic papers, each using different models and formats. Translating these parameters into executable scripts remains a time-consuming and error-prone process. To improve efficiency in physics research and accelerate the cosmological simulation process, we introduce SimAgents, a multi-agent system designed to automate both parameter configuration from the literature and preliminary analysis for cosmology research. SimAgents is powered by specialized LLM agents capable of physics reasoning, simulation software validation, and tool execution. These agents collaborate through structured communication, ensuring that extracted parameters are physically meaningful, internally consistent, and software-compliant. We also construct a cosmological parameter extraction evaluation dataset by collecting over 40 simulations in published papers from Arxiv and leading journals that cover diverse simulation types. Experiments on the dataset demonstrate a strong performance of SimAgents, highlighting its effectiveness and potential to accelerate scientific research for physicists. Our demonstration video is available at: https://youtu.be/w1zLpm_CaWA. The complete system and dataset are publicly available at https://github.com/xwzhang98/SimAgents.

Miaojing Shi, Xiaowen Zhang, Zijie Yue et al.

Recent advances in large vision-language models (VLMs) have shown remarkable progress in solving the text-promptable object counting problem. Representative methods typically specify text prompts with object category information in images. This however is insufficient for training the model to accurately distinguish the number of objects in the counting task. To this end, we propose QUANet, which introduces novel quantity-oriented text prompts with a vision-text quantity alignment loss to enhance the model's quantity awareness. Moreover, we propose a dual-stream adaptive counting decoder consisting of a Transformer stream, a CNN stream, and a number of Transformer-to-CNN enhancement adapters (T2C-adapters) for density map prediction. The T2C-adapters facilitate the effective knowledge communication and aggregation between the Transformer and CNN streams. A cross-stream quantity ranking loss is proposed in the end to optimize the ranking orders of predictions from the two streams. Extensive experiments on standard benchmarks such as FSC-147, CARPK, PUCPR+, and ShanghaiTech demonstrate our model's strong generalizability for zero-shot class-agnostic counting. Code is available at https://github.com/viscom-tongji/QUANet

Yide Liu, Haijiang Sun, Xiaowen Zhang et al.

Remote Sensing Image Super-Resolution (RSISR) reconstructs high-resolution (HR) remote sensing images from low-resolution inputs to support fine-grained ground object interpretation. Existing methods face three key challenges: (1) Difficulty in extracting multi-scale features from spatially heterogeneous RS scenes, (2) Limited prior information causing semantic inconsistency in reconstructions, and (3) Trade-off imbalance between geometric accuracy and visual quality. To address these issues, we propose the Texture Transfer Residual Denoising Dual Diffusion Model (TTRD3) with three innovations: First, a Multi-scale Feature Aggregation Block (MFAB) employing parallel heterogeneous convolutional kernels for multi-scale feature extraction. Second, a Sparse Texture Transfer Guidance (STTG) module that transfers HR texture priors from reference images of similar scenes. Third, a Residual Denoising Dual Diffusion Model (RDDM) framework combining residual diffusion for deterministic reconstruction and noise diffusion for diverse generation. Experiments on multi-source RS datasets demonstrate TTRD3's superiority over state-of-the-art methods, achieving 1.43% LPIPS improvement and 3.67% FID enhancement compared to best-performing baselines. Code/model: https://github.com/LED-666/TTRD3.

Shin Hwei Tan, Chunfeng Hu, Ziqiang Li et al.

Many studies have shown the benefits of introducing open-source projects into teaching Software Engineering (SE) courses. However, there are several limitations of existing studies that limit the wide adaptation of open-source projects in a classroom setting, including (1) the selected project is limited to one particular project, (2) most studies only investigated on its effect on teaching a specific SE concept, and (3) students may make mistakes in their contribution which leads to poor quality code. Meanwhile, software companies have successfully launched programs like Google Summer of Code (GSoC) and FindBugs "fixit" to contribute to open-source projects. Inspired by the success of these programs, we propose GitHub-OSS Fixit, a course project where students are taught to contribute to open-source Java projects by fixing bugs reported in GitHub. We described our course outline to teach students SE concepts by encouraging the usages of several automated program analysis tools. We also included the carefully designed instructions that we gave to students for participating in GitHub-OSS Fixit. As all lectures and labs are conducted online, we think that our course design could help in guiding future online SE courses. Overall, our survey results show that students think that GitHub-OSS Fixit could help them to improve many skills and apply the knowledge taught in class. In total, 154 students have submitted 214 pull requests to 24 different Java projects, in which 59 of them have been merged, and 82 have been closed by developers.

Xiaowen Zhang, Ziming Zhou, Fengnian Zhao et al.

Deep learning surrogates for CFD flow-field prediction often rely on large, complex models, which can be slow and fragile when data are noisy or incomplete. We introduce FlowForge, a staged local rollout engine that predicts future flow fields by compiling a locality-preserving update schedule and executing it with a shared lightweight local predictor. Rather than producing the next frame in a single global pass, FlowForge rewrites spatial sites stage by stage so that each update conditions only on bounded local context exposed by earlier stages. This compile-execute design aligns inference with short-range physical dependence, keeps latency predictable, and limits error amplification from global mixing. Across PDEBench, CFDBench, and BubbleML, FlowForge matches or improves upon strong baselines in pointwise accuracy, delivers consistently better robustness to noise and missing observations, and maintains stable multi-step rollout behavior while reducing per-step latency.

Xintong Zhang, Zhi Gao, Bofei Zhang et al.

Vision language models (VLMs) have achieved impressive performance across a variety of computer vision tasks. However, the multimodal reasoning capability has not been fully explored in existing models. In this paper, we propose a Chain-of-Focus (CoF) method that allows VLMs to perform adaptive focusing and zooming in on key image regions based on obtained visual cues and the given questions, achieving efficient multimodal reasoning. To enable this CoF capability, we present a two-stage training pipeline, including supervised fine-tuning (SFT) and reinforcement learning (RL). In the SFT stage, we construct the MM-CoF dataset, comprising 3K samples derived from a visual agent designed to adaptively identify key regions to solve visual tasks with different image resolutions and questions. We use MM-CoF to fine-tune the Qwen2.5-VL model for cold start. In the RL stage, we leverage the outcome accuracies and formats as rewards to update the Qwen2.5-VL model, enabling further refining the search and reasoning strategy of models without human priors. Our model achieves significant improvements on multiple benchmarks. On the V* benchmark that requires strong visual reasoning capability, our model outperforms existing VLMs by 5% among 8 image resolutions ranging from 224 to 4K, demonstrating the effectiveness of the proposed CoF method and facilitating the more efficient deployment of VLMs in practical applications.

Dihan Zheng, Yihang Zou, Xiaowen Zhang et al.

The data bottleneck has emerged as a fundamental challenge in learning based image restoration methods. Researchers have attempted to generate synthesized training data using paired or unpaired samples to address this challenge. This study proposes SeNM-VAE, a semi-supervised noise modeling method that leverages both paired and unpaired datasets to generate realistic degraded data. Our approach is based on modeling the conditional distribution of degraded and clean images with a specially designed graphical model. Under the variational inference framework, we develop an objective function for handling both paired and unpaired data. We employ our method to generate paired training samples for real-world image denoising and super-resolution tasks. Our approach excels in the quality of synthetic degraded images compared to other unpaired and paired noise modeling methods. Furthermore, our approach demonstrates remarkable performance in downstream image restoration tasks, even with limited paired data. With more paired data, our method achieves the best performance on the SIDD dataset.

Xiaowen Zhang, Hannuo Zhang, Shin Hwei Tan

Modern agentic frameworks (e.g., CrewAI and AutoGen) have evolved into complex, autonomous multi-agent systems, introducing unique reliability challenges beyond earlier pipeline-based LLM libraries. However, existing empirical studies focus on earlier LLM libraries or task-level bugs, leaving the unique complexities of these agentic frameworks unexplored. We bridge the gap by conducting a comprehensive study of 409 fixed bugs from five representative agentic frameworks. We propose a five-layer abstraction to capture structural complexities in agentic frameworks, spanning from orchestration to infrastructure. Our study uncovers specialized symptoms, such as unexpected execution sequences and user configurations ignored, which are unique to autonomous orchestration. We further identify agent-specific root causes, including modelrelated faults, cognitive context mismanagement, and orchestration faults. Statistical analysis reveals cross-framework consistency and significant associations among these bug dimensions. Finally, our automated pattern mining identifies frequent bug-triggering patterns (e.g., model backend-ID combinations), and we show their transferability across different framework designs. Our findings facilitate cross-platform testing and improve the reliability of agentic systems.

Xiaochen Zhao, Kaikai Wang, Xiaowen Zhang et al.

Large language model (LLM) agents demonstrate strong performance in short-text contexts but often underperform in extended dialogues due to inefficient memory management. Existing approaches face a fundamental trade-off between efficiency and effectiveness: memory compression risks losing critical details required for complex reasoning, while retaining raw text introduces unnecessary computational overhead for simple queries. The crux lies in the limitations of monolithic memory representations and static retrieval mechanisms, which fail to emulate the flexible and proactive memory scheduling capabilities observed in humans, thus struggling to adapt to diverse problem scenarios. Inspired by the principle of cognitive economy, we propose HyMem, a hybrid memory architecture that enables dynamic on-demand scheduling through multi-granular memory representations. HyMem adopts a dual-granular storage scheme paired with a dynamic two-tier retrieval system: a lightweight module constructs summary-level context for efficient response generation, while an LLM-based deep module is selectively activated only for complex queries, augmented by a reflection mechanism for iterative reasoning refinement. Experiments show that HyMem achieves strong performance on both the LOCOMO and LongMemEval benchmarks, outperforming full-context while reducing computational cost by 92.6\%, establishing a state-of-the-art balance between efficiency and performance in long-term memory management.

Xintong Zhang, Xiaowen Zhang, Jongrong Wu et al.

Adaptive multimodal reasoning has emerged as a promising frontier in Vision-Language Models (VLMs), aiming to dynamically modulate between tool-augmented visual reasoning and text reasoning to enhance both effectiveness and efficiency. However, existing evaluations rely on static difficulty labels and simplistic metrics, which fail to capture the dynamic nature of difficulty relative to varying model capacities. Consequently, they obscure the distinction between adaptive mode selection and general performance while neglecting fine-grained process analyses. In this paper, we propose AdaptMMBench, a comprehensive benchmark for adaptive multimodal reasoning across five domains: real-world, OCR, GUI, knowledge, and math, encompassing both direct perception and complex reasoning tasks. AdaptMMBench utilizes a Matthews Correlation Coefficient (MCC) metric to evaluate the selection rationality of different reasoning modes, isolating this meta-cognition ability by dynamically identifying task difficulties based on models' capability boundaries. Moreover, AdaptMMBench facilitates multi-dimensional process evaluation across key step coverage, tool effectiveness, and computational efficiency. Our evaluation reveals that while adaptive mode selection scales with model capacity, it notably decouples from final accuracy. Conversely, key step coverage aligns with performance, though tool effectiveness remains highly inconsistent across model architectures.

Negar Hassanpour, Muhammad Kamran Janjua, Kunlin Zhang et al.

Balancing competing objectives remains a fundamental challenge in multi-task learning (MTL), primarily due to conflicting gradients across individual tasks. A common solution relies on computing a dynamic gradient update vector that balances competing tasks as optimization progresses. Building on this idea, we propose ConicGrad, a principled, scalable, and robust MTL approach formulated as a constrained optimization problem. Our method introduces an angular constraint to dynamically regulate gradient update directions, confining them within a cone centered on the reference gradient of the overall objective. By balancing task-specific gradients without over-constraining their direction or magnitude, ConicGrad effectively resolves inter-task gradient conflicts. Moreover, our framework ensures computational efficiency and scalability to high-dimensional parameter spaces. We conduct extensive experiments on standard supervised learning and reinforcement learning MTL benchmarks, and demonstrate that ConicGrad achieves state-of-the-art performance across diverse tasks.

Yuntao Du, Yinghao Chen, Fengli Cui et al.

Unsupervised domain adaptation aims to transfer knowledge from a labeled source domain to an unlabeled target domain. Previous methods focus on learning domain-invariant features to decrease the discrepancy between the feature distributions as well as minimizing the source error and have made remarkable progress. However, a recently proposed theory reveals that such a strategy is not sufficient for a successful domain adaptation. It shows that besides a small source error, both the discrepancy between the feature distributions and the discrepancy between the labeling functions should be small across domains. The discrepancy between the labeling functions is essentially the cross-domain errors which are ignored by existing methods. To overcome this issue, in this paper, a novel method is proposed to integrate all the objectives into a unified optimization framework. Moreover, the incorrect pseudo labels widely used in previous methods can lead to error accumulation during learning. To alleviate this problem, the pseudo labels are obtained by utilizing structural information of the target domain besides source classifier and we propose a curriculum learning based strategy to select the target samples with more accurate pseudo-labels during training. Comprehensive experiments are conducted, and the results validate that our approach outperforms state-of-the-art methods.

Yuntao Du, Ruiting Zhang, Xiaowen Zhang et al.

Although achieving remarkable progress, it is very difficult to induce a supervised classifier without any labeled data. Unsupervised domain adaptation is able to overcome this challenge by transferring knowledge from a labeled source domain to an unlabeled target domain. Transferability and discriminability are two key criteria for characterizing the superiority of feature representations to enable successful domain adaptation. In this paper, a novel method called \textit{learning TransFerable and Discriminative Features for unsupervised domain adaptation} (TFDF) is proposed to optimize these two objectives simultaneously. On the one hand, distribution alignment is performed to reduce domain discrepancy and learn more transferable representations. Instead of adopting \textit{Maximum Mean Discrepancy} (MMD) which only captures the first-order statistical information to measure distribution discrepancy, we adopt a recently proposed statistic called \textit{Maximum Mean and Covariance Discrepancy} (MMCD), which can not only capture the first-order statistical information but also capture the second-order statistical information in the reproducing kernel Hilbert space (RKHS). On the other hand, we propose to explore both local discriminative information via manifold regularization and global discriminative information via minimizing the proposed \textit{class confusion} objective to learn more discriminative features, respectively. We integrate these two objectives into the \textit{Structural Risk Minimization} (RSM) framework and learn a domain-invariant classifier. Comprehensive experiments are conducted on five real-world datasets and the results verify the effectiveness of the proposed method.

Yuntao Du, Zhiwen Tan, Qian Chen et al.

Unsupervised domain adaptation aims at transferring knowledge from the labeled source domain to the unlabeled target domain. Previous adversarial domain adaptation methods mostly adopt the discriminator with binary or $K$-dimensional output to perform marginal or conditional alignment independently. Recent experiments have shown that when the discriminator is provided with domain information in both domains and label information in the source domain, it is able to preserve the complex multimodal information and high semantic information in both domains. Following this idea, we adopt a discriminator with $2K$-dimensional output to perform both domain-level and class-level alignments simultaneously in a single discriminator. However, a single discriminator can not capture all the useful information across domains and the relationships between the examples and the decision boundary are rarely explored before. Inspired by multi-view learning and latest advances in domain adaptation, besides the adversarial process between the discriminator and the feature extractor, we also design a novel mechanism to make two discriminators pit against each other, so that they can provide diverse information for each other and avoid generating target features outside the support of the source domain. To the best of our knowledge, it is the first time to explore a dual adversarial strategy in domain adaptation. Moreover, we also use the semi-supervised learning regularization to make the representations more discriminative. Comprehensive experiments on two real-world datasets verify that our method outperforms several state-of-the-art domain adaptation methods.