Wenyu Zhang

Ke Liu, Jiwei Wei, Wenyu Zhang et al.

With rapid advances in audio-visual generative models, reliable forgery detection becomes increasingly critical. Existing methods for audio-visual deepfake detection typically rely on cross-modal inconsistencies. In singing, rhythmic vocalization weakens this coupling and introduces a nontrivial domain shift, substantially degrading detection performance. We construct the Singing Head DeepFake (SHDF) dataset using rhythm-aware generative models to fill the gap in singing benchmarks. To cope with cross-scenario domain shifts, we propose a Text-guided Audio-Visual Forgery Detection (T-AVFD) framework that generalizes across both talking and singing scenarios. T-AVFD comprises a facial authenticity pattern learner and a multi-modal differential weight learning module. The pattern learner aligns facial features with multi-granularity textual descriptions to learn generalizable authenticity patterns. The weight learning module preserves intrinsic audio-visual consistency and adaptively integrates it with authenticity patterns via differential weighting. Extensive experiments on multiple talking head deepfake datasets and SHDF show consistent improvements over existing baselines and strong robustness under diverse perturbations.

Joy Dhar, Song Xia, Manish Kumar Pandey et al.

We introduce Hybrid Convolutions with Attention Stochasticity (HyCAS), an adversarial defense that narrows the long-standing gap between provable robustness under L2 certificates and empirical robustness against strong L attacks, while preserving strong generalization across diverse imaging benchmarks. HyCAS unifies deterministic and randomized principles by coupling 1-Lipschitz, spectrally normalized convolutions with two stochastic components, spectral normalized random, projection filters and a randomized attention-noise mechanism, to realize a randomized defense. Injecting smoothing randomness inside the architecture yields an overall <= 2-Lipschitz network with formal certificates. Exten-sive experiments on diverse imaging benchmarks, including CIFAR-10/100, ImageNet-1k, NIH Chest X-ray, HAM10000, show that HyCAS surpasses prior leading certified and empirical defenses, boosting certified accuracy by up to 7.3% (on NIH Chest X-ray) and empirical robustness by up to 3.1% (on HAM10000), without sacrificing clean accuracy. These results show that a randomized Lipschitz constrained architecture can simultaneously improve both certified L2 and empirical L adversarial robustness, thereby supporting safer deployment of deep models in high-stakes applications. Code: https://github.com/misti1203/HyCAS

Jiyuan Xu, Wenyu Zhang, Xin Jing et al.

Current methods for multivariate time series forecasting can be classified into channel-dependent and channel-independent models. Channel-dependent models learn cross-channel features but often overfit the channel ordering, which hampers adaptation when channels are added or reordered. Channel-independent models treat each channel in isolation to increase flexibility, yet this neglects inter-channel dependencies and limits performance. To address these limitations, we propose \textbf{CPiRi}, a \textbf{channel permutation invariant (CPI)} framework that infers cross-channel structure from data rather than memorizing a fixed ordering, enabling deployment in settings with structural and distributional co-drift without retraining. CPiRi couples \textbf{spatio-temporal decoupling architecture} with \textbf{permutation-invariant regularization training strategy}: a frozen pretrained temporal encoder extracts high-quality temporal features, a lightweight spatial module learns content-driven inter-channel relations, while a channel shuffling strategy enforces CPI during training. We further \textbf{ground CPiRi in theory} by analyzing permutation equivariance in multivariate time series forecasting. Experiments on multiple benchmarks show state-of-the-art results. CPiRi remains stable when channel orders are shuffled and exhibits strong \textbf{inductive generalization} to unseen channels even when trained on \textbf{only half} of the channels, while maintaining \textbf{practical efficiency} on large-scale datasets. The source code is released at https://github.com/JasonStraka/CPiRi.

Wenyu Zhang, Xin Deng, Baojun Jia et al.

Current Scene text image super-resolution approaches primarily focus on extracting robust features, acquiring text information, and complex training strategies to generate super-resolution images. However, the upsampling module, which is crucial in the process of converting low-resolution images to high-resolution ones, has received little attention in existing works. To address this issue, we propose the Pixel Adapter Module (PAM) based on graph attention to address pixel distortion caused by upsampling. The PAM effectively captures local structural information by allowing each pixel to interact with its neighbors and update features. Unlike previous graph attention mechanisms, our approach achieves 2-3 orders of magnitude improvement in efficiency and memory utilization by eliminating the dependency on sparse adjacency matrices and introducing a sliding window approach for efficient parallel computation. Additionally, we introduce the MLP-based Sequential Residual Block (MSRB) for robust feature extraction from text images, and a Local Contour Awareness loss ($\mathcal{L}_{lca}$) to enhance the model's perception of details. Comprehensive experiments on TextZoom demonstrate that our proposed method generates high-quality super-resolution images, surpassing existing methods in recognition accuracy. For single-stage and multi-stage strategies, we achieved improvements of 0.7\% and 2.6\%, respectively, increasing the performance from 52.6\% and 53.7\% to 53.3\% and 56.3\%. The code is available at https://github.com/wenyu1009/RTSRN.

Wenyu Zhang, Qing Ding, Jian Hu et al.

Graph convolutional networks (GCN) is widely used to handle irregular data since it updates node features by using the structure information of graph. With the help of iterated GCN, high-order information can be obtained to further enhance the representation of nodes. However, how to apply GCN to structured data (such as pictures) has not been deeply studied. In this paper, we explore the application of graph attention networks (GAT) in image feature extraction. First of all, we propose a novel graph generation algorithm to convert images into graphs through matrix transformation. It is one magnitude faster than the algorithm based on K Nearest Neighbors (KNN). Then, GAT is used on the generated graph to update the node features. Thus, a more robust representation is obtained. These two steps are combined into a module called pixel-wise graph attention module (PGA). Since the graph obtained by our graph generation algorithm can still be transformed into a picture after processing, PGA can be well combined with CNN. Based on these two modules, we consulted the ResNet and design a pixel-wise graph attention network (PGANet). The PGANet is applied to the task of person re-identification in the datasets Market1501, DukeMTMC-reID and Occluded-DukeMTMC (outperforms state-of-the-art by 0.8\%, 1.1\% and 11\% respectively, in mAP scores). Experiment results show that it achieves the state-of-the-art performance. \href{https://github.com/wenyu1009/PGANet}{The code is available here}.

Wenyu Zhang, Li Shen, Chuan-Sheng Foo

Source-free domain adaptation (SFDA) aims to adapt a source model trained on a fully-labeled source domain to an unlabeled target domain. Large-data pre-trained networks are used to initialize source models during source training, and subsequently discarded. However, source training can cause the model to overfit to source data distribution and lose applicable target domain knowledge. We propose to integrate the pre-trained network into the target adaptation process as it has diversified features important for generalization and provides an alternate view of features and classification decisions different from the source model. We propose to distil useful target domain information through a co-learning strategy to improve target pseudolabel quality for finetuning the source model. Evaluation on 4 benchmark datasets show that our proposed strategy improves adaptation performance and can be successfully integrated with existing SFDA methods. Leveraging modern pre-trained networks that have stronger representation learning ability in the co-learning strategy further boosts performance.

Wenyu Zhang, Li Shen, Wanyue Zhang et al.

Deep networks are prone to performance degradation when there is a domain shift between the source (training) data and target (test) data. Recent test-time adaptation methods update batch normalization layers of pre-trained source models deployed in new target environments with streaming data to mitigate such performance degradation. Although such methods can adapt on-the-fly without first collecting a large target domain dataset, their performance is dependent on streaming conditions such as mini-batch size and class-distribution, which can be unpredictable in practice. In this work, we propose a framework for few-shot domain adaptation to address the practical challenges of data-efficient adaptation. Specifically, we propose a constrained optimization of feature normalization statistics in pre-trained source models supervised by a small support set from the target domain. Our method is easy to implement and improves source model performance with as few as one sample per class for classification tasks. Extensive experiments on 5 cross-domain classification and 4 semantic segmentation datasets show that our method achieves more accurate and reliable performance than test-time adaptation, while not being constrained by streaming conditions.

Khoi Nguyen Tiet Nguyen, Wenyu Zhang, Kangkang Lu et al.

Deep learning models achieve remarkable accuracy in computer vision tasks, yet remain vulnerable to adversarial examples--carefully crafted perturbations to input images that can deceive these models into making confident but incorrect predictions. This vulnerability pose significant risks in high-stakes applications such as autonomous vehicles, security surveillance, and safety-critical inspection systems. While the existing literature extensively covers adversarial attacks in image classification, comprehensive analyses of such attacks on object detection systems remain limited. This paper presents a novel taxonomic framework for categorizing adversarial attacks specific to object detection architectures, synthesizes existing robustness metrics, and provides a comprehensive empirical evaluation of state-of-the-art attack methodologies on popular object detection models, including both traditional detectors and modern detectors with vision-language pretraining. Through rigorous analysis of open-source attack implementations and their effectiveness across diverse detection architectures, we derive key insights into attack characteristics. Furthermore, we delineate critical research gaps and emerging challenges to guide future investigations in securing object detection systems against adversarial threats. Our findings establish a foundation for developing more robust detection models while highlighting the urgent need for standardized evaluation protocols in this rapidly evolving domain.

Wenyu Zhang, Mohamed Ragab, Chuan-Sheng Foo

Domain generalization methods aim to learn models robust to domain shift with data from a limited number of source domains and without access to target domain samples during training. Popular domain alignment methods for domain generalization seek to extract domain-invariant features by minimizing the discrepancy between feature distributions across all domains, disregarding inter-domain relationships. In this paper, we instead propose a novel representation learning methodology that selectively enforces prediction consistency between source domains estimated to be closely-related. Specifically, we hypothesize that domains share different class-informative representations, so instead of aligning all domains which can cause negative transfer, we only regularize the discrepancy between closely-related domains. We apply our method to time-series classification tasks and conduct comprehensive experiments on three public real-world datasets. Our method significantly improves over the baseline and achieves better or competitive performance in comparison with state-of-the-art methods in terms of both accuracy and model calibration.

Xin Deng, WenYu Zhang, Qing Ding et al.

In point cloud analysis, point-based methods have rapidly developed in recent years. These methods have recently focused on concise MLP structures, such as PointNeXt, which have demonstrated competitiveness with Convolutional and Transformer structures. However, standard MLPs are limited in their ability to extract local features effectively. To address this limitation, we propose a Vector-oriented Point Set Abstraction that can aggregate neighboring features through higher-dimensional vectors. To facilitate network optimization, we construct a transformation from scalar to vector using independent angles based on 3D vector rotations. Finally, we develop a PointVector model that follows the structure of PointNeXt. Our experimental results demonstrate that PointVector achieves state-of-the-art performance $\textbf{72.3\% mIOU}$ on the S3DIS Area 5 and $\textbf{78.4\% mIOU}$ on the S3DIS (6-fold cross-validation) with only $\textbf{58\%}$ model parameters of PointNeXt. We hope our work will help the exploration of concise and effective feature representations. The code will be released soon.

Mingzhe Lu, Yiwen Wang, Yanbing Liu et al.

Computational narrative analysis aims to capture rhythm, tension, and emotional dynamics in literary texts. Existing large language models can generate long stories but overly focus on causal coherence, neglecting the complex story arcs and orchestration inherent in human narratives. This suggests a structural misalignment between model- and human-generated narratives. We therefore position narrative analysis as a diagnostic proxy for generation and propose VISTA Space, a high-dimensional framework for narrative orchestration that unifies human and model perspectives while jointly characterizing narrative function and structure in a common space. We further introduce LitVISTA, a structurally annotated benchmark grounded in literary texts, which operationalizes VISTA Space for systematic evaluation of models' narrative orchestration capabilities. Under an oracle setting with gold event anchors, we evaluate frontier LLMs including GPT, Claude, Grok, and Gemini. Results reveal systematic deficiencies, as current models struggle to jointly capture narrative function and structure and fail to form an integrated global view of literary narrative orchestration. End-to-end analysis further shows that failures are dominated by anchor identification and localization errors. Even advanced thinking modes yield mixed and often limited gains for literary narrative understanding.

Wenyu Zhang, Guoliang You, Tianlun et al.

Coding agents are increasingly used as application builders, yet many evaluations still focus on source code, repository-level tests, or intermediate traces rather than the delivered application. We introduce WebGameBench, a requirement-to-application benchmark that evaluates whether coding agents can turn a frozen Structured WebGame Specification into a browser-accessible game. Browser-native games provide a compact but behavior-dense testbed: even simple games require coordinated input handling, spatial mapping, rule execution, state transitions, terminal conditions, restart behavior, and visible feedback. In WebGameBench, each generated artifact is built, served, and exposed as a browser-accessible application under a unified deployment protocol. A runtime evaluator then interacts with the delivered game in a real browser and assigns a three-way label: EXCELLENT, USABLE, or UNUSABLE. On a human-reviewed subset, the runtime label is broadly aligned with human gameplay review under the Usable-rate criterion. Across 111 tasks, 12 coding agents, and 14 evaluation configurations, WebGameBench separates current systems: the best configuration reaches a 76.9% usable rate but only a 20.2% excellent rate. This gap shows that crossing the minimum playable-delivery threshold is still far from complete requirement satisfaction. To our knowledge, WebGameBench is the first requirement-to-application benchmark for browser-native game delivery that validates delivered-application runtime labels against independent human gameplay review under the Usable-rate criterion.

Wenyu Zhang, Li Shen, Chuan-Sheng Foo

Source-free domain adaptation (SFDA) aims to adapt a source model trained on a fully-labeled source domain to a related but unlabeled target domain. While the source model is a key avenue for acquiring target pseudolabels, the generated pseudolabels may exhibit source bias. In the conventional SFDA pipeline, a large data (e.g. ImageNet) pre-trained feature extractor is used to initialize the source model at the start of source training, and subsequently discarded. Despite having diverse features important for generalization, the pre-trained feature extractor can overfit to the source data distribution during source training and forget relevant target domain knowledge. Rather than discarding this valuable knowledge, we introduce an integrated framework to incorporate pre-trained networks into the target adaptation process. The proposed framework is flexible and allows us to plug modern pre-trained networks into the adaptation process to leverage their stronger representation learning capabilities. For adaptation, we propose the Co-learn algorithm to improve target pseudolabel quality collaboratively through the source model and a pre-trained feature extractor. Building on the recent success of the vision-language model CLIP in zero-shot image recognition, we present an extension Co-learn++ to further incorporate CLIP's zero-shot classification decisions. We evaluate on 4 benchmark datasets and include more challenging scenarios such as open-set, partial-set and open-partial SFDA. Experimental results demonstrate that our proposed strategy improves adaptation performance and can be successfully integrated with existing SFDA methods. Project code is available at https://github.com/zwenyu/colearn-plus.

Guoliang You, Xiaomeng Chu, Yifan Duan et al.

When planning for autonomous driving, it is crucial to consider essential traffic elements such as lanes, intersections, traffic regulations, and dynamic agents. However, they are often overlooked by the traditional end-to-end planning methods, likely leading to inefficiencies and non-compliance with traffic regulations. In this work, we endeavor to integrate the perception of these elements into the planning task. To this end, we propose Perception Helps Planning (PHP), a novel framework that reconciles lane-level planning with perception. This integration ensures that planning is inherently aligned with traffic constraints, thus facilitating safe and efficient driving. Specifically, PHP focuses on both edges of a lane for planning and perception purposes, taking into consideration the 3D positions of both lane edges and attributes for lane intersections, lane directions, lane occupancy, and planning. In the algorithmic design, the process begins with the transformer encoding multi-camera images to extract the above features and predicting lane-level perception results. Next, the hierarchical feature early fusion module refines the features for predicting planning attributes. Finally, the double-edge interpreter utilizes a late-fusion process specifically designed to integrate lane-level perception and planning information, culminating in the generation of vehicle control signals. Experiments on three Carla benchmarks show significant improvements in driving score of 27.20%, 33.47%, and 15.54% over existing algorithms, respectively, achieving the state-of-the-art performance, with the system operating up to 22.57 FPS.

Wenyu Zhang, Shuo Sun, Bin Wang et al.

The rapid advancements in large language models (LLMs) have significantly enhanced natural language processing capabilities, facilitating the development of AudioLLMs that process and understand speech and audio inputs alongside text. Existing AudioLLMs typically combine a pre-trained audio encoder with a pre-trained LLM, which are subsequently finetuned on specific audio tasks. However, the pre-trained audio encoder has constrained capacity to capture features for new tasks and datasets. To address this, we propose to incorporate mixtures of `weak' encoders (MoWE) into the AudioLLM framework. MoWE supplements a base encoder with a pool of relatively light weight encoders, selectively activated based on the audio input to enhance feature extraction without significantly increasing model size. Our empirical results demonstrate that MoWE effectively improves multi-task performance, broadening the applicability of AudioLLMs to more diverse audio tasks.

Wenyu Zhang, Wei En Ng, Lixin Ma et al.

Current vision-language models may grasp basic spatial cues and simple directions (e.g. left, right, front, back), but struggle with the multi-dimensional spatial reasoning necessary for human-like understanding and real-world applications. To address this gap, we develop SPHERE (Spatial Perception and Hierarchical Evaluation of REasoning), a hierarchical evaluation framework supported by a new human-annotated dataset. SPHERE systematically probes models across increasing levels of complexity, from fundamental skills to multi-skill integration and high-level reasoning that combines spatial, visual, and logical understanding. Benchmark evaluation of state-of-the-art models reveals significant deficiencies, especially in reasoning about distance and proximity, understanding both egocentric and allocentric perspectives, and applying spatial logic in physical contexts. These findings expose critical blind spots in existing models and underscore the need for more advanced spatial reasoning techniques, driving the development of vision-language models that align more closely with human spatial cognition. The SPHERE benchmark is available at https://github.com/zwenyu/SPHERE-VLM.

Daxiang Dong, Mingming Zheng, Dong Xu et al.

We present Qianfan-OCR, a 4B-parameter end-to-end vision-language model that unifies document parsing, layout analysis, and document understanding within a single architecture. It performs direct image-to-Markdown conversion and supports diverse prompt-driven tasks including table extraction, chart understanding, document QA, and key information extraction. To address the loss of explicit layout analysis in end-to-end OCR, we propose Layout-as-Thought, an optional thinking phase triggered by special think tokens that generates structured layout representations -- bounding boxes, element types, and reading order -- before producing final outputs, recovering layout grounding capabilities while improving accuracy on complex layouts. Qianfan-OCR ranks first among end-to-end models on OmniDocBench v1.5 (93.12) and OlmOCR Bench (79.8), achieves competitive results on OCRBench, CCOCR, DocVQA, and ChartQA against general VLMs of comparable scale, and attains the highest average score on public key information extraction benchmarks, surpassing Gemini-3.1-Pro, Seed-2.0, and Qwen3-VL-235B. The model is publicly accessible via the Baidu AI Cloud Qianfan platform.

Daxiang Dong, Mingming Zheng, Dong Xu et al.

We present Qianfan-VL, a series of multimodal large language models ranging from 3B to 70B parameters, achieving state-of-the-art performance through innovative domain enhancement techniques. Our approach employs multi-stage progressive training and high-precision data synthesis pipelines, which prove to be critical technologies for enhancing domain-specific capabilities while maintaining strong general performance. Qianfan-VL achieves comparable results to leading open-source models on general benchmarks, with state-of-the-art performance on benchmarks such as CCBench, SEEDBench IMG, ScienceQA, and MMStar. The domain enhancement strategy delivers significant advantages in OCR and document understanding, validated on both public benchmarks (OCRBench 873, DocVQA 94.75%) and in-house evaluations. Notably, Qianfan-VL-8B and 70B variants incorporate long chain-of-thought capabilities, demonstrating superior performance on mathematical reasoning (MathVista 78.6%) and logical inference tasks. All models are trained entirely on Baidu's Kunlun P800 chips, validating the capability of large-scale AI infrastructure to train SOTA-level multimodal models with over 90% scaling efficiency on 5000 chips for a single task. This work establishes an effective methodology for developing domain-enhanced multimodal models suitable for diverse enterprise deployment scenarios.

Bin Wang, Xunlong Zou, Geyu Lin et al.

We introduce AudioBench, a universal benchmark designed to evaluate Audio Large Language Models (AudioLLMs). It encompasses 8 distinct tasks and 26 datasets, among which, 7 are newly proposed datasets. The evaluation targets three main aspects: speech understanding, audio scene understanding, and voice understanding (paralinguistic). Despite recent advancements, there lacks a comprehensive benchmark for AudioLLMs on instruction following capabilities conditioned on audio signals. AudioBench addresses this gap by setting up datasets as well as desired evaluation metrics. Besides, we also evaluated the capabilities of five popular models and found that no single model excels consistently across all tasks. We outline the research outlook for AudioLLMs and anticipate that our open-sourced evaluation toolkit, data, and leaderboard will offer a robust testbed for future model developments.

Skyler Seto, Martin T. Wells, Wenyu Zhang

Deep neural networks achieve state-of-the-art performance in a variety of tasks by extracting a rich set of features from unstructured data, however this performance is closely tied to model size. Modern techniques for inducing sparsity and reducing model size are (1) network pruning, (2) training with a sparsity inducing penalty, and (3) training a binary mask jointly with the weights of the network. We study different sparsity inducing penalties from the perspective of Bayesian hierarchical models and present a novel penalty called Hierarchical Adaptive Lasso (HALO) which learns to adaptively sparsify weights of a given network via trainable parameters. When used to train over-parametrized networks, our penalty yields small subnetworks with high accuracy without fine-tuning. Empirically, on image recognition tasks, we find that HALO is able to learn highly sparse network (only 5% of the parameters) with significant gains in performance over state-of-the-art magnitude pruning methods at the same level of sparsity. Code is available at https://github.com/skyler120/sparsity-halo.

Tianshu Zhu, Wenyu Zhang, Xiaoying Zuo et al.

SWE-bench-style agentic reinforcement learning relies on expensive stateful trajectories, yet substantial compute is wasted on sampled rollout groups with skewed pass rates, where binary rewards provide a weak contrastive signal. We frame this inefficiency as a pass-rate control problem and show that a 50% pass rate is the most informative operating point: it maximizes reward entropy, the probability of surviving group filtering, RLOO advantage energy under GRPO, and success--failure contrastive structure. Guided by this principle, we propose Prefix Sampling (PS), which replays trajectory prefixes to steer skewed groups toward this regime: successful prefixes serve as head starts for mostly failing groups, while failing prefixes serve as handicaps for mostly passing groups. In stateful agent environments, prefix states are reconstructed through replay while replayed tokens are excluded from the loss, restricting optimization to continuations generated by the current policy. On SWE-bench-style agentic RL, PS delivers end-to-end wall-clock speedups of 2.01x on Qwen3-14B and 1.55x on Qwen3-32B while preserving or improving final verified performance. For 14B, the SWE-bench Verified peak rises from the baseline peak of 0.273 to 0.295 under PS. Additional mathematical reasoning experiments on AIME 2025 show the same pass-rate control pattern and decompose the gains into replay, bidirectional coverage, and adaptive control.

Aritra Roy, Kevin Shen, Andrew MacBride et al.

Large language models (LLMs) are rapidly changing how researchers in materials science and chemistry discover, organize, and act on scientific knowledge. This paper analyzes a broad set of community-developed LLM applications in an effort to identify emerging patterns in how these systems can be used across the scientific research lifecycle. We organize the projects into two complementary categories: Knowledge Infrastructure, systems that structure, retrieve, synthesize, and validate scientific information; and Action Systems, systems that execute, coordinate, or automate scientific work across computational and experimental environments. The submissions reveal a shift from single-purpose LLM tools toward integrated, multi-agent workflows that combine retrieval, reasoning, tool use, and domain-specific validation. Prominent themes include retrieval-augmented generation as grounding infrastructure, persistent structured knowledge representations, multimodal and multilingual scientific inputs, and early progress toward laboratory-integrated closed-loop systems. Together, these results suggest that LLMs are evolving from general-purpose assistants into composable infrastructure for scientific reasoning and action. This work provides a community snapshot of that transition and a practical taxonomy for understanding emerging LLM-enabled workflows in materials science and chemistry.

Haotian Zhao, Yuxin Zhang, Songlin Zhou et al.

Reinforcement learning (RL) has significantly advanced the ability of large language model (LLM) agents to interact with environments and solve multi-turn tasks. Yet effective training remains challenging, as sparse, outcome-only rewards make it difficult to assign credit to individual steps in an agent's action trajectory. A common remedy is to introduce dense intermediate supervision, such as process reward models or auxiliary self-supervised signals, but this increases supervision and tuning complexity and often generalizes poorly across tasks and domains. This paper presents AEM, a supervision-free credit assignment method that adaptively modulates entropy dynamics during RL training to achieve a more effective exploration-exploitation trade-off. Theoretically, we elevate entropy analysis from the token level to the response level to reduce token sampling variance and show that entropy drift under natural gradients is intrinsically governed by the product of the advantage and the relative response surprisal. Specifically, we derive a practical proxy to reshape training dynamics, enabling a natural transition from exploration to exploitation. Extensive experiments across various benchmarks and models ranging from 1.5B to 32B parameters demonstrate the effectiveness of AEM, including a notable 1.4 percent gain when integrated into a state-of-the-art baseline on the highly challenging SWE-bench-Verified benchmark.

Yingxu He, Zhuohan Liu, Shuo Sun et al.

We introduce MERaLiON-AudioLLM (Multimodal Empathetic Reasoning and Learning in One Network), the first speech-text model tailored for Singapore's multilingual and multicultural landscape. Developed under the National Large Language Models Funding Initiative, Singapore, MERaLiON-AudioLLM integrates advanced speech and text processing to address the diverse linguistic nuances of local accents and dialects, enhancing accessibility and usability in complex, multilingual environments. Our results demonstrate improvements in both speech recognition and task-specific understanding, positioning MERaLiON-AudioLLM as a pioneering solution for region specific AI applications. We envision this release to set a precedent for future models designed to address localised linguistic and cultural contexts in a global framework.

Bin Wang, Xunlong Zou, Shuo Sun et al.

Singlish, a Creole language rooted in English, is a key focus in linguistic research within multilingual and multicultural contexts. However, its spoken form remains underexplored, limiting insights into its linguistic structure and applications. To address this gap, we standardize and annotate the largest spoken Singlish corpus, introducing the Multitask National Speech Corpus (MNSC). These datasets support diverse tasks, including Automatic Speech Recognition (ASR), Spoken Question Answering (SQA), Spoken Dialogue Summarization (SDS), and Paralinguistic Question Answering (PQA). We release standardized splits and a human-verified test set to facilitate further research. Additionally, we propose SingAudioLLM, a multi-task multimodal model leveraging multimodal large language models to handle these tasks concurrently. Experiments reveal our models adaptability to Singlish context, achieving state-of-the-art performance and outperforming prior models by 10-30% in comparison with other AudioLLMs and cascaded solutions.

Wenyu Zhang, Qingmu Liu, Felix Ong Wei Cong et al.

Domain adaptation is a critical task in machine learning that aims to improve model performance on a target domain by leveraging knowledge from a related source domain. In this work, we introduce Universal Semi-Supervised Domain Adaptation (UniSSDA), a practical yet challenging setting where the target domain is partially labeled, and the source and target label space may not strictly match. UniSSDA is at the intersection of Universal Domain Adaptation (UniDA) and Semi-Supervised Domain Adaptation (SSDA): the UniDA setting does not allow for fine-grained categorization of target private classes not represented in the source domain, while SSDA focuses on the restricted closed-set setting where source and target label spaces match exactly. Existing UniDA and SSDA methods are susceptible to common-class bias in UniSSDA settings, where models overfit to data distributions of classes common to both domains at the expense of private classes. We propose a new prior-guided pseudo-label refinement strategy to reduce the reinforcement of common-class bias due to pseudo-labeling, a common label propagation strategy in domain adaptation. We demonstrate the effectiveness of the proposed strategy on benchmark datasets Office-Home, DomainNet, and VisDA. The proposed strategy attains the best performance across UniSSDA adaptation settings and establishes a new baseline for UniSSDA.

Qiongqiong Wang, Hardik Bhupendra Sailor, Tianchi Liu et al.

Recent speech-LLMs have shown impressive performance in tasks like transcription and translation, yet they remain limited in understanding the paralinguistic aspects of speech crucial for social and emotional intelligence. We propose CP-Bench, a benchmark for evaluating speech-LLMs on contextual paralinguistic reasoning the integration of verbal content with non-verbal cues like emotion and prosody. The benchmark includes two curated question answering (QA) datasets requiring both linguistic and empathetic understanding. We evaluate state-of-the-art speech-LLMs from both open and closed-source models and perform a comprehensive analysis across different question types. The top two models were further analyzed under temperature tuning to understand its effect on this task. Our benchmark reveals a key gap in existing evaluations and offers insights into building more context-aware and emotionally intelligent speech-capable LLMs.

Qiongqiong Wang, Hardik B. Sailor, Jeremy H. M. Wong et al.

Current large speech language models (Speech-LLMs) often exhibit limitations in empathetic reasoning, primarily due to the absence of training datasets that integrate both contextual content and paralinguistic cues. In this work, we propose two approaches to incorporate contextual paralinguistic information into model training: (1) an explicit method that provides paralinguistic metadata (e.g., emotion annotations) directly to the LLM, and (2) an implicit method that automatically generates novel training question-answer (QA) pairs using both categorical and dimensional emotion annotations alongside speech transcriptions. Our implicit method boosts performance (LLM-judged) by 38.41% on a human-annotated QA benchmark, reaching 46.02% when combined with the explicit approach, showing effectiveness in contextual paralinguistic understanding. We also validate the LLM judge by demonstrating its correlation with classification metrics, providing support for its reliability.

Rui Cao, Wei Tu, Dongsheng Chen et al.

The shift toward high-quality urbanization has brought increased attention to the issue of "urban villages", which has become a prominent social problem in China. However, there is a lack of available geospatial data on urban villages, making it crucial to prioritize urban village mapping. In order to assess the current progress in urban village mapping and identify challenges and future directions, we have conducted a comprehensive review, which to the best of our knowledge is the first of its kind in this field. Our review begins by providing a clear context for urban villages and elaborating the method for literature review, then summarizes the study areas, data sources, and approaches used for urban village mapping in China. We also address the challenges and future directions for further research. Through thorough investigation, we find that current studies only cover very limited study areas and periods and lack sufficient investigation into the scalability, transferability, and interpretability of identification approaches due to the challenges in concept fuzziness and variances, spatial heterogeneity and variances of urban villages, and data availability. Future research can complement and further the current research in the following potential directions in order to achieve large-area mapping across the whole nation...

Wenjing Gong, Xinyue Ye, Keshu Wu et al.

Extreme heat events, exacerbated by climate change, pose significant challenges to urban resilience and planning. This study introduces a climate-responsive digital twin framework integrating the Spatiotemporal Vision Transformer (ST-ViT) model to enhance heat stress forecasting and decision-making. Using a Texas campus as a testbed, we synthesized high-resolution physical model simulations with spatial and meteorological data to develop fine-scale human thermal predictions. The ST-ViT-powered digital twin enables efficient, data-driven insights for planners and stakeholders, supporting targeted heat mitigation strategies and advancing climate-adaptive urban design. This campus-scale demonstration offers a foundation for future applications across broader and more diverse urban contexts.

Wenyu Zhang, Yao Tong, Yiqiu Liu et al.

Urban villages (UVs), informal settlements embedded within China's urban fabric, have undergone widespread demolition and redevelopment in recent decades. However, there remains a lack of systematic evaluation of whether the demolished land has been effectively reused, raising concerns about the efficacy and sustainability of current redevelopment practices. To address the gap, this study proposes a deep learning-based framework to monitor the spatiotemporal changes of UVs in China. Specifically, semantic segmentation of multi-temporal remote sensing imagery is first used to map evolving UV boundaries, and then post-demolition land use is classified into six categories based on the "remained-demolished-redeveloped" phase: incomplete demolition, vacant land, construction sites, buildings, green spaces, and others. Four representative cities from China's four economic regions were selected as the study areas, i.e., Guangzhou (East), Zhengzhou (Central), Xi'an (West), and Harbin (Northeast). The results indicate: 1) UV redevelopment processes were frequently prolonged; 2) redevelopment transitions primarily occurred in peripheral areas, whereas urban cores remained relatively stable; and 3) three spatiotemporal transformation pathways, i.e., synchronized redevelopment, delayed redevelopment, and gradual optimization, were revealed. This study highlights the fragmented, complex and nonlinear nature of UV redevelopment, underscoring the need for tiered and context-sensitive planning strategies. By linking spatial dynamics with the context of redevelopment policies, the findings offer valuable empirical insights that support more inclusive, efficient, and sustainable urban renewal, while also contributing to a broader global understanding of informal settlement transformations.

Wenyu Zhang, Jie Luo, Xinming Zhang et al.

With the explosive growth of multimodal content online, pre-trained visual-language models have shown great potential for multimodal recommendation. However, while these models achieve decent performance when applied in a frozen manner, surprisingly, due to significant domain gaps (e.g., feature distribution discrepancy and task objective misalignment) between pre-training and personalized recommendation, adopting a joint training approach instead leads to performance worse than baseline. Existing approaches either rely on simple feature extraction or require computationally expensive full model fine-tuning, struggling to balance effectiveness and efficiency. To tackle these challenges, we propose \textbf{P}arameter-efficient \textbf{T}uning for \textbf{M}ultimodal \textbf{Rec}ommendation (\textbf{PTMRec}), a novel framework that bridges the domain gap between pre-trained models and recommendation systems through a knowledge-guided dual-stage parameter-efficient training strategy. This framework not only eliminates the need for costly additional pre-training but also flexibly accommodates various parameter-efficient tuning methods.

Gaofeng Chen, Yaoduo Zhang, Li Huang et al.

Supervised deep-learning (SDL) techniques with paired training datasets have been widely studied for X-ray computed tomography (CT) image reconstruction. However, due to the difficulties of obtaining paired training datasets in clinical routine, the SDL methods are still away from common uses in clinical practices. In recent years, self-supervised deep-learning (SSDL) techniques have shown great potential for the studies of CT image reconstruction. In this work, we propose a self-supervised cross-task mutual learning (SS-CTML) framework for CT image reconstruction. Specifically, a sparse-view scanned and a limited-view scanned sinogram data are first extracted from a full-view scanned sinogram data, which results in three individual reconstruction tasks, i.e., the full-view CT (FVCT) reconstruction, the sparse-view CT (SVCT) reconstruction, and limited-view CT (LVCT) reconstruction. Then, three neural networks are constructed for the three reconstruction tasks. Considering that the ultimate goals of the three tasks are all to reconstruct high-quality CT images, we therefore construct a set of cross-task mutual learning objectives for the three tasks, in which way, the three neural networks can be self-supervised optimized by learning from each other. Clinical datasets are adopted to evaluate the effectiveness of the proposed framework. Experimental results demonstrate that the SS-CTML framework can obtain promising CT image reconstruction performance in terms of both quantitative and qualitative measurements.

Holy Lovenia, Rahmad Mahendra, Salsabil Maulana Akbar et al.

Southeast Asia (SEA) is a region rich in linguistic diversity and cultural variety, with over 1,300 indigenous languages and a population of 671 million people. However, prevailing AI models suffer from a significant lack of representation of texts, images, and audio datasets from SEA, compromising the quality of AI models for SEA languages. Evaluating models for SEA languages is challenging due to the scarcity of high-quality datasets, compounded by the dominance of English training data, raising concerns about potential cultural misrepresentation. To address these challenges, we introduce SEACrowd, a collaborative initiative that consolidates a comprehensive resource hub that fills the resource gap by providing standardized corpora in nearly 1,000 SEA languages across three modalities. Through our SEACrowd benchmarks, we assess the quality of AI models on 36 indigenous languages across 13 tasks, offering valuable insights into the current AI landscape in SEA. Furthermore, we propose strategies to facilitate greater AI advancements, maximizing potential utility and resource equity for the future of AI in SEA.

Mohamed Ragab, Peiliang Gong, Emadeldeen Eldele et al.

Source-free domain adaptation (SFDA) aims to adapt a model pre-trained on a labeled source domain to an unlabeled target domain without access to source data, preserving the source domain's privacy. While SFDA is prevalent in computer vision, it remains largely unexplored in time series analysis. Existing SFDA methods, designed for visual data, struggle to capture the inherent temporal dynamics of time series, hindering adaptation performance. This paper proposes MAsk And imPUte (MAPU), a novel and effective approach for time series SFDA. MAPU addresses the critical challenge of temporal consistency by introducing a novel temporal imputation task. This task involves randomly masking time series signals and leveraging a dedicated temporal imputer to recover the original signal within the learned embedding space, bypassing the complexities of noisy raw data. Notably, MAPU is the first method to explicitly address temporal consistency in the context of time series SFDA. Additionally, it offers seamless integration with existing SFDA methods, providing greater flexibility. We further introduce E-MAPU, which incorporates evidential uncertainty estimation to address the overconfidence issue inherent in softmax predictions. To achieve that, we leverage evidential deep learning to obtain a better-calibrated pre-trained model and adapt the target encoder to map out-of-support target samples to a new feature representation closer to the source domain's support. This fosters better alignment, ultimately enhancing adaptation performance. Extensive experiments on five real-world time series datasets demonstrate that both MAPU and E-MAPU achieve significant performance gains compared to existing methods. These results highlight the effectiveness of our proposed approaches for tackling various time series domain adaptation problems.

Astha Garg, Wenyu Zhang, Jules Samaran et al.

Several techniques for multivariate time series anomaly detection have been proposed recently, but a systematic comparison on a common set of datasets and metrics is lacking. This paper presents a systematic and comprehensive evaluation of unsupervised and semi-supervised deep-learning based methods for anomaly detection and diagnosis on multivariate time series data from cyberphysical systems. Unlike previous works, we vary the model and post-processing of model errors, i.e. the scoring functions independently of each other, through a grid of 10 models and 4 scoring functions, comparing these variants to state of the art methods. In time-series anomaly detection, detecting anomalous events is more important than detecting individual anomalous time-points. Through experiments, we find that the existing evaluation metrics either do not take events into account, or cannot distinguish between a good detector and trivial detectors, such as a random or an all-positive detector. We propose a new metric to overcome these drawbacks, namely, the composite F-score ($Fc_1$), for evaluating time-series anomaly detection. Our study highlights that dynamic scoring functions work much better than static ones for multivariate time series anomaly detection, and the choice of scoring functions often matters more than the choice of the underlying model. We also find that a simple, channel-wise model - the Univariate Fully-Connected Auto-Encoder, with the dynamic Gaussian scoring function emerges as a winning candidate for both anomaly detection and diagnosis, beating state of the art algorithms.

Wenyu Zhang, Florian Meyer

Seamless situational awareness provided by modern radar systems relies on effective methods for multiobject tracking (MOT). This paper presents a graph-based Bayesian method for nonlinear and high-dimensional MOT problems that embeds particle flow. To perform operations on the graph effectively, particles are migrated towards regions of high likelihood based on the solution of a partial differential equation. This makes it possible to obtain good object detection and tracking performance with a relatively small number of particles even if object states are high dimensional and sensor measurements are very informative. Simulation results demonstrate reduced computational complexity and memory requirements as well as favorable detection and estimation accuracy in a challenging 3-D MOT scenario.

Wenyu Zhang

Online prediction for streaming time series data has practical use for many real-world applications where downstream decisions depend on accurate forecasts for the future. Deployment in dynamic environments requires models to adapt quickly to changing data distributions without overfitting. We propose POLA (Predicting Online by Learning rate Adaptation) to automatically regulate the learning rate of recurrent neural network models to adapt to changing time series patterns across time. POLA meta-learns the learning rate of the stochastic gradient descent (SGD) algorithm by assimilating the prequential or interleaved-test-then-train evaluation scheme for online prediction. We evaluate POLA on two real-world datasets across three commonly-used recurrent neural network models. POLA demonstrates overall comparable or better predictive performance over other online prediction methods.

Wenyu Zhang, Mohamed Ragab, Ramon Sagarna

While deep neural networks demonstrate state-of-the-art performance on a variety of learning tasks, their performance relies on the assumption that train and test distributions are the same, which may not hold in real-world applications. Domain generalization addresses this issue by employing multiple source domains to build robust models that can generalize to unseen target domains subject to shifts in data distribution. In this paper, we propose Domain-Free Domain Generalization (DFDG), a model-agnostic method to achieve better generalization performance on the unseen test domain without the need for source domain labels. DFDG uses novel strategies to learn domain-invariant class-discriminative features. It aligns class relationships of samples through class-conditional soft labels, and uses saliency maps, traditionally developed for post-hoc analysis of image classification networks, to remove superficial observations from training inputs. DFDG obtains competitive performance on both time series sensor and image classification public datasets.

An Nguyen, Wenyu Zhang, Leo Schwinn et al.

Process Mining has recently gained popularity in healthcare due to its potential to provide a transparent, objective and data-based view on processes. Conformance checking is a sub-discipline of process mining that has the potential to answer how the actual process executions deviate from existing guidelines. In this work, we analyze a medical training process for a surgical procedure. Ten students were trained to install a Central Venous Catheters (CVC) with ultrasound. Event log data was collected directly after instruction by the supervisors during a first test run and additionally after a subsequent individual training phase. In order to provide objective performance measures, we formulate an optimal, global sequence alignment problem inspired by approaches in bioinformatics. Therefore, we use the Petri net model representation of the medical process guideline to simulate a representative set of guideline conform sequences. Next, we calculate the optimal, global sequence alignment of the recorded and simulated event logs. Finally, the output measures and visualization of aligned sequences are provided for objective feedback.

Wenyu Zhang, Skyler Seto, Devesh K. Jha

Learning accurate models of the physical world is required for a lot of robotic manipulation tasks. However, during manipulation, robots are expected to interact with unknown workpieces so that building predictive models which can generalize over a number of these objects is highly desirable. In this paper, we study the problem of designing deep learning agents which can generalize their models of the physical world by building context-aware learning models. The purpose of these agents is to quickly adapt and/or generalize their notion of physics of interaction in the real world based on certain features about the interacting objects that provide different contexts to the predictive models. With this motivation, we present context-aware zero shot learning (CAZSL, pronounced as casual) models, an approach utilizing a Siamese network architecture, embedding space masking and regularization based on context variables which allows us to learn a model that can generalize to different parameters or features of the interacting objects. We test our proposed learning algorithm on the recently released Omnipush datatset that allows testing of meta-learning capabilities using low-dimensional data. Codes for CAZSL are available at https://www.merl.com/research/license/CAZSL.

Wenyu Zhang, Devesh K. Jha, Emil Laftchiev et al.

This paper addresses a multi-label predictive fault classification problem for multidimensional time-series data. While fault (event) detection problems have been thoroughly studied in literature, most of the state-of-the-art techniques can't reliably predict faults (events) over a desired future horizon. In the most general setting of these types of problems, one or more samples of data across multiple time series can be assigned several concurrent fault labels from a finite, known set and the task is to predict the possibility of fault occurrence over a desired time horizon. This type of problem is usually accompanied by strong class imbalances where some classes are represented by only a few samples. Importantly, in many applications of the problem such as fault prediction and predictive maintenance, it is exactly these rare classes that are of most interest. To address the problem, this paper proposes a general approach that utilizes a multi-label recurrent neural network with a new cost function that accentuates learning in the imbalanced classes. The proposed algorithm is tested on two public benchmark datasets: an industrial plant dataset from the PHM Society Data Challenge, and a human activity recognition dataset. The results are compared with state-of-the-art techniques for time-series classification and evaluation is performed using the F1-score, precision and recall.

Wenyu Zhang, Daniel Gilbert, David Matteson

Change detection involves segmenting sequential data such that observations in the same segment share some desired properties. Multivariate change detection continues to be a challenging problem due to the variety of ways change points can be correlated across channels and the potentially poor signal-to-noise ratio on individual channels. In this paper, we are interested in locating additive outliers (AO) and level shifts (LS) in the unsupervised setting. We propose ABACUS, Automatic BAyesian Changepoints Under Sparsity, a Bayesian source separation technique to recover latent signals while also detecting changes in model parameters. Multi-level sparsity achieves both dimension reduction and modeling of signal changes. We show ABACUS has competitive or superior performance in simulation studies against state-of-the-art change detection methods and established latent variable models. We also illustrate ABACUS on two real application, modeling genomic profiles and analyzing household electricity consumption.

Skyler Seto, Wenyu Zhang, Yichen Zhou

Accurate and computationally efficient means for classifying human activities have been the subject of extensive research efforts. Most current research focuses on extracting complex features to achieve high classification accuracy. We propose a template selection approach based on Dynamic Time Warping, such that complex feature extraction and domain knowledge is avoided. We demonstrate the predictive capability of the algorithm on both simulated and real smartphone data.