arXiv subjects

Wei-Yao Wang

Publications and source records attributed to Wei-Yao Wang.

At least 19 recordsLinked to original sources

arXiv2026

Omni-Interactive Universal Embedder

Multimodal representation learning has been shifting from traditional two-tower architectures to large language model (LLM)-based embedders due to their strong instruction-following capabilities. Despite this progress, existing approaches primarily focus on language and image modalities, which also remain the dominant modalities for user-conditioned interactions in current embedders. In this paper, we propose the first Omni-Interactive Universal Embedder (OmniUE), which not only learns a unified embedding space across text, video, and audio by leveraging intermediate-layer representations from dedicated learnable tokens, but also supports omni-interactive querying, enabling users to provide inputs in the form of text, visual regions of interest, and audio spans. Within OmniUE, visual and audio segmenters process diverse user interactions and integrate them with an omni-LLM to produce user-conditioned any-to-any embeddings via context aggregation. To evaluate OmniUE's omni-interactive capabilities, we introduce OmniCHOIR, benchmarking models for omni-interactive compositional audio retrieval based on the given text, video, and audio as well as unimodal or multimodal interaction prompts. OmniUE consistently surpasses state-of-the-art baselines across diverse modalities, with average improvements of 10.5% on textual-interactive video benchmarks (MMEB-v2-video), 1.1% on audio tasks (MAEB), 83.7% on visual-interactive benchmarks (SCaR), and 24.1% on our omni-interactive OmniCHOIR benchmark. We believe that jointly advancing omni-modal representation learning and omni-interactive querying paves the way toward universal embedders.

cs.AI↗

arXiv2026

Benchmarking Agentic Newswriting via Journalistic Workflows

Recent advances in autonomous digital agents from industry (e.g., Manus AI and Gemini's research mode) highlight their potential for structured tasks through autonomous decision-making and task decomposition, but it remains unclear how well such systems support real-world information-intensive workflows. We study this question in journalism, where newswriting requires iterative planning, contextual reasoning, and active discovery of missing background to produce a coherent article. We introduce NEWSAGENT, a benchmark for evaluating how agents search raw materials, select relevant information, and iteratively revise drafts through core journalistic functions. Given a writing instruction and partial firsthand materials, agents must identify narrative perspectives, issue keyword-based queries, retrieve historical context, and generate complete news articles. Unlike typical summarization or retrieval tasks, essential context is not directly available and must be actively discovered, reflecting real-world reporting constraints. NEWSAGENT consists of 6k human-verified examples derived from real news. We evaluate open- and closed-sourced LLMs with commonly-used agentic frameworks on NEWSAGENT, which shows that agents are capable of retrieving relevant facts but struggling with planning and narrative integration. We believe that NEWSAGENT serves a realistic testbed for iterating and evaluating agent capabilities in terms of web data manipulation to real-world productivity.

cs.AI↗

arXiv2026

Seeing is Understanding: Unlocking Causal Attention into Modality-Mutual Attention for Multimodal LLMs

Recent Multimodal Large Language Models (MLLMs) have demonstrated significant progress in perceiving and reasoning over multimodal inquiries, ushering in a new research era for foundation models. However, vision-language misalignment in MLLMs has emerged as a critical challenge, where the textual responses generated by these models are not factually aligned with the given text-image inputs. Existing efforts to address vision-language misalignment have focused on developing specialized vision-language connectors or leveraging visual instruction tuning from diverse domains. In this paper, we tackle this issue from a fundamental yet unexplored perspective by revisiting the core architecture of MLLMs. Most MLLMs are typically built on decoder-only LLMs consisting of a causal attention mechanism, which limits the ability of the earlier modalities (e.g., images) to incorporate information from the latter modalities (e.g., text). To address this problem a MLLM that unlocks causal attention into our proposed modality-mutual attention (MMA) to enable image tokens to attend to text tokens. This simple yet effective design allows MMA to achieve state-of-the-art performance in 12 multimodal understanding benchmarks (+6.2% on average across 3 LLMs backbones) without introducing additional parameters. Our MMA design is intended to be generic, allowing for applications across various modalities, and scalable to accommodate diverse multimodal scenarios.

cs.CV↗

arXiv2026

Echoes Over Time: Unlocking Length Generalization in Video-to-Audio Generation Models

Scaling multimodal alignment between video and audio is challenging, particularly due to limited data and the mismatch between text descriptions and frame-level video information. In this work, we tackle the scaling challenge in multimodal-to-audio generation, examining whether models trained on short instances can generalize to longer ones during testing. To tackle this challenge, we present multimodal hierarchical networks so-called MMHNet, an enhanced extension of state-of-the-art video-to-audio models. Our approach integrates a hierarchical method and non-causal Mamba to support long-form audio generation. Our proposed method significantly improves long audio generation up to more than 5 minutes. We also prove that training short and testing long is possible in the video-to-audio generation tasks without training on the longer durations. We show in our experiments that our proposed method could achieve remarkable results on long-video to audio benchmarks, beating prior works in video-to-audio tasks. Moreover, we showcase our model capability in generating more than 5 minutes, while prior video-to-audio methods fall short in generating with long durations.

cs.CV↗

arXiv2026

SF-Mamba: Rethinking State Space Model for Vision

The realm of Mamba for vision has been advanced in recent years to strike for the alternatives of Vision Transformers (ViTs) that suffer from the quadratic complexity. While the recurrent scanning mechanism of Mamba offers computational efficiency, it inherently limits non-causal interactions between image patches. Prior works have attempted to address this limitation through various multi-scan strategies; however, these approaches suffer from inefficiencies due to suboptimal scan designs and frequent data rearrangement. Moreover, Mamba exhibits relatively slow computational speed under short token lengths, commonly used in visual tasks. In pursuit of a truly efficient vision encoder, we rethink the scan operation for vision and the computational efficiency of Mamba. To this end, we propose SF-Mamba, a novel visual Mamba with two key proposals: auxiliary patch swapping for encoding bidirectional information flow under an unidirectional scan and batch folding with periodic state reset for advanced GPU parallelism. Extensive experiments on image classification, object detection, and instance and semantic segmentation consistently demonstrate that our proposed SF-Mamba significantly outperforms state-of-the-art baselines while improving throughput across different model sizes. We will release the source code after publication.

cs.CV↗

arXiv2026

VIRTUE: Visual-Interactive Text-Image Universal Embedder

Multimodal representation learning models have demonstrated successful operation across complex tasks, and the integration of vision-language models (VLMs) has further enabled embedding models with instruction-following capabilities. However, existing embedding models lack visual-interactive capabilities to specify regions of interest from users (e.g., point, bounding box, mask), which have been explored in generative models to broaden their human-interactive applicability. Equipping embedding models with visual interactions not only would unlock new applications with localized grounding of user intent, which remains unexplored, but also enable the models to learn entity-level information within images to complement their global representations for conventional embedding tasks. In this paper, we propose a novel Visual-InteRactive Text-Image Universal Embedder (VIRTUE) that extends the capabilities of the segmentation model and the vision-language model to the realm of representation learning. In VIRTUE, the segmentation model can process visual prompts that pinpoint specific regions within an image, thereby enabling the embedder to handle complex and ambiguous scenarios more precisely. To evaluate the visual-interaction ability of VIRTUE, we introduce a large-scale Segmentation-and-Scene Caption Retrieval (SCaR) benchmark comprising 1M samples that aims to retrieve the text caption by jointly considering the entity with a specific object and image scene. VIRTUE consistently achieves a state-of-the-art performance with significant improvements across 36 universal MMEB (3.1%-8.5%) and five visual-interactive SCaR (15.2%-20.3%) tasks.

cs.AI↗

arXiv2025

WebGen-V Bench: Structured Representation for Enhancing Visual Design in LLM-based Web Generation and Evaluation

Witnessed by the recent advancements on leveraging LLM for coding and multimodal understanding, we present WebGen-V, a new benchmark and framework for instruction-to-HTML generation that enhances both data quality and evaluation granularity. WebGen-V contributes three key innovations: (1) an unbounded and extensible agentic crawling framework that continuously collects real-world webpages and can leveraged to augment existing benchmarks; (2) a structured, section-wise data representation that integrates metadata, localized UI screenshots, and JSON-formatted text and image assets, explicit alignment between content, layout, and visual components for detailed multimodal supervision; and (3) a section-level multimodal evaluation protocol aligning text, layout, and visuals for high-granularity assessment. Experiments with state-of-the-art LLMs and ablation studies validate the effectiveness of our structured data and section-wise evaluation, as well as the contribution of each component. To the best of our knowledge, WebGen-V is the first work to enable high-granularity agentic crawling and evaluation for instruction-to-HTML generation, providing a unified pipeline from real-world data acquisition and webpage generation to structured multimodal assessment.

cs.AI↗

arXiv2025

MCA: Modality Composition Awareness for Robust Composed Multimodal Retrieval

Multimodal retrieval, which seeks to retrieve relevant content across modalities such as text or image, supports applications from AI search to contents production. Despite the success of separate-encoder approaches like CLIP align modality-specific embeddings with contrastive learning, recent multimodal large language models (MLLMs) enable a unified encoder that directly processes composed inputs. While flexible and advanced, we identify that unified encoders trained with conventional contrastive learning are prone to learn modality shortcut, leading to poor robustness under distribution shifts. We propose a modality composition awareness framework to mitigate this issue. Concretely, a preference loss enforces multimodal embeddings to outperform their unimodal counterparts, while a composition regularization objective aligns multimodal embeddings with prototypes composed from its unimodal parts. These objectives explicitly model structural relationships between the composed representation and its unimodal counterparts. Experiments on various benchmarks show gains in out-of-distribution retrieval, highlighting modality composition awareness as a effective principle for robust composed multimodal retrieval when utilizing MLLMs as the unified encoder.

cs.CL↗

arXiv2025

Mixture Experts with Test-Time Self-Supervised Aggregation for Tabular Imbalanced Regression

Tabular data serve as a fundamental and ubiquitous representation of structured information in numerous real-world applications, e.g., finance and urban planning. In the realm of tabular imbalanced applications, data imbalance has been investigated in classification tasks with insufficient instances in certain labels, causing the model's ineffective generalizability. However, the imbalance issue of tabular regression tasks is underexplored, and yet is critical due to unclear boundaries for continuous labels and simplifying assumptions in existing imbalance regression work, which often rely on known and balanced test distributions. Such assumptions may not hold in practice and can lead to performance degradation. To address these issues, we propose MATI: Mixture Experts with Test-Time Self-Supervised Aggregation for Tabular Imbalance Regression, featuring two key innovations: (i) the Region-Aware Mixture Expert, which adopts a Gaussian Mixture Model to capture the underlying related regions. The statistical information of each Gaussian component is then used to synthesize and train region-specific experts to capture the unique characteristics of their respective regions. (ii) Test-Time Self-Supervised Expert Aggregation, which dynamically adjusts region expert weights based on test data features to reinforce expert adaptation across varying test distributions. We evaluated MATI on four real-world tabular imbalance regression datasets, including house pricing, bike sharing, and age prediction. To reflect realistic deployment scenarios, we adopted three types of test distributions: a balanced distribution with uniform target frequencies, a normal distribution that follows the training data, and an inverse distribution that emphasizes rare target regions. On average across these three test distributions, MATI achieved a 7.1% improvement in MAE compared to existing methods.

cs.LG↗

arXiv2025

LLM4TS: Aligning Pre-Trained LLMs as Data-Efficient Time-Series Forecasters

Multivariate time-series forecasting is vital in various domains, e.g., economic planning and weather prediction. Deep train-from-scratch models have exhibited effective performance yet require large amounts of data, which limits real-world applicability. Recently, researchers have leveraged the representation learning transferability of pre-trained Large Language Models (LLMs) to handle limited non-linguistic datasets effectively. However, incorporating LLMs with time-series data presents challenges of limited adaptation due to different compositions between time-series and linguistic data, and the inability to process multi-scale temporal information. To tackle these challenges, we propose LLM4TS, a framework for time-series forecasting with pre-trained LLMs. LLM4TS consists of a two-stage fine-tuning strategy: the time-series alignment stage to align LLMs with the nuances of time-series data, and the forecasting fine-tuning stage for downstream time-series forecasting tasks. Furthermore, our framework features a novel two-level aggregation method that integrates multi-scale temporal data within pre-trained LLMs, enhancing their ability to interpret time-specific information. In experiments across 7 time-series forecasting datasets, LLM4TS is superior to existing state-of-the-art methods compared with trained-from-scratch models in full-shot scenarios, and also achieves the highest rank in few-shot scenarios. In addition, evaluations compared with different unsupervised representation learning approaches highlight LLM4TS's effectiveness with representation learning in forecasting tasks. Ablation studies further validate each component's contribution to LLM4TS and underscore the essential role of utilizing LLM's pre-trained weights for optimal performance. The code is available at https://github.com/blacksnail789521/LLM4TS.

cs.LG↗

arXiv2025

Talk Structurally, Act Hierarchically: A Collaborative Framework for LLM Multi-Agent Systems

Recent advancements in LLM-based multi-agent (LLM-MA) systems have shown promise, yet significant challenges remain in managing communication and refinement when agents collaborate on complex tasks. In this paper, we propose \textit{Talk Structurally, Act Hierarchically (TalkHier)}, a novel framework that introduces a structured communication protocol for context-rich exchanges and a hierarchical refinement system to address issues such as incorrect outputs, falsehoods, and biases. \textit{TalkHier} surpasses various types of SoTA, including inference scaling model (OpenAI-o1), open-source multi-agent models (e.g., AgentVerse), and majority voting strategies on current LLM and single-agent baselines (e.g., ReAct, GPT4o), across diverse tasks, including open-domain question answering, domain-specific selective questioning, and practical advertisement text generation. These results highlight its potential to set a new standard for LLM-MA systems, paving the way for more effective, adaptable, and collaborative multi-agent frameworks. The code is available https://github.com/sony/talkhier.

cs.AI↗

arXiv2024

APAR: Modeling Irregular Target Functions in Tabular Regression via Arithmetic-Aware Pre-Training and Adaptive-Regularized Fine-Tuning

Tabular data are fundamental in common machine learning applications, ranging from finance to genomics and healthcare. This paper focuses on tabular regression tasks, a field where deep learning (DL) methods are not consistently superior to machine learning (ML) models due to the challenges posed by irregular target functions inherent in tabular data, causing sensitive label changes with minor variations from features. To address these issues, we propose a novel Arithmetic-Aware Pre-training and Adaptive-Regularized Fine-tuning framework (APAR), which enables the model to fit irregular target function in tabular data while reducing the negative impact of overfitting. In the pre-training phase, APAR introduces an arithmetic-aware pretext objective to capture intricate sample-wise relationships from the perspective of continuous labels. In the fine-tuning phase, a consistency-based adaptive regularization technique is proposed to self-learn appropriate data augmentation. Extensive experiments across 10 datasets demonstrated that APAR outperforms existing GBDT-, supervised NN-, and pretrain-finetune NN-based methods in RMSE (+9.43% $\sim$ 20.37%), and empirically validated the effects of pre-training tasks, including the study of arithmetic operations. Our code and data are publicly available at https://github.com/johnnyhwu/APAR.

cs.LG↗

arXiv2024

Self-Supervised Learning of Disentangled Representations for Multivariate Time-Series

Multivariate time-series data in fields like healthcare and industry are informative but challenging due to high dimensionality and lack of labels. Recent self-supervised learning methods excel in learning rich representations without labels but struggle with disentangled embeddings and inductive bias issues like transformation-invariance. To address these challenges, we introduce TimeDRL, a framework for multivariate time-series representation learning with dual-level disentangled embeddings. TimeDRL features: (i) disentangled timestamp-level and instance-level embeddings using a [CLS] token strategy; (ii) timestamp-predictive and instance-contrastive tasks for representation learning; and (iii) avoidance of augmentation methods to eliminate inductive biases. Experiments on forecasting and classification datasets show TimeDRL outperforms existing methods, with further validation in semi-supervised settings with limited labeled data.

cs.LG↗

arXiv2024

A Survey on Self-Supervised Learning for Non-Sequential Tabular Data

Self-supervised learning (SSL) has been incorporated into many state-of-the-art models in various domains, where SSL defines pretext tasks based on unlabeled datasets to learn contextualized and robust representations. Recently, SSL has become a new trend in exploring the representation learning capability in the realm of tabular data, which is more challenging due to not having explicit relations for learning descriptive representations. This survey aims to systematically review and summarize the recent progress and challenges of SSL for non-sequential tabular data (SSL4NS-TD). We first present a formal definition of NS-TD and clarify its correlation to related studies. Then, these approaches are categorized into three groups - predictive learning, contrastive learning, and hybrid learning, with their motivations and strengths of representative methods in each direction. Moreover, application issues of SSL4NS-TD are presented, including automatic data engineering, cross-table transferability, and domain knowledge integration. In addition, we elaborate on existing benchmarks and datasets for NS-TD applications to analyze the performance of existing tabular models. Finally, we discuss the challenges of SSL4NS-TD and provide potential directions for future research. We expect our work to be useful in terms of encouraging more research on lowering the barrier to entry SSL for the tabular domain, and of improving the foundations for implicit tabular data.

cs.LG↗

arXiv2024

Offline Imitation of Badminton Player Behavior via Experiential Contexts and Brownian Motion

In the dynamic and rapid tactic involvements of turn-based sports, badminton stands out as an intrinsic paradigm that requires alter-dependent decision-making of players. While the advancement of learning from offline expert data in sequential decision-making has been witnessed in various domains, how to rally-wise imitate the behaviors of human players from offline badminton matches has remained underexplored. Replicating opponents' behavior benefits players by allowing them to undergo strategic development with direction before matches. However, directly applying existing methods suffers from the inherent hierarchy of the match and the compounding effect due to the turn-based nature of players alternatively taking actions. In this paper, we propose RallyNet, a novel hierarchical offline imitation learning model for badminton player behaviors: (i) RallyNet captures players' decision dependencies by modeling decision-making processes as a contextual Markov decision process. (ii) RallyNet leverages the experience to generate context as the agent's intent in the rally. (iii) To generate more realistic behavior, RallyNet leverages Geometric Brownian Motion (GBM) to model the interactions between players by introducing a valuable inductive bias for learning player behaviors. In this manner, RallyNet links player intents with interaction models with GBM, providing an understanding of interactions for sports analytics. We extensively validate RallyNet with the largest available real-world badminton dataset consisting of men's and women's singles, demonstrating its ability to imitate player behaviors. Results reveal RallyNet's superiority over offline imitation learning methods and state-of-the-art turn-based approaches, outperforming them by at least 16% in mean rule-based agent normalization score. Furthermore, we discuss various practical use cases to highlight RallyNet's applicability.

cs.AI↗

arXiv2024

TimeDRL: Disentangled Representation Learning for Multivariate Time-Series

Multivariate time-series data in numerous real-world applications (e.g., healthcare and industry) are informative but challenging due to the lack of labels and high dimensionality. Recent studies in self-supervised learning have shown their potential in learning rich representations without relying on labels, yet they fall short in learning disentangled embeddings and addressing issues of inductive bias (e.g., transformation-invariance). To tackle these challenges, we propose TimeDRL, a generic multivariate time-series representation learning framework with disentangled dual-level embeddings. TimeDRL is characterized by three novel features: (i) disentangled derivation of timestamp-level and instance-level embeddings from patched time-series data using a [CLS] token strategy; (ii) utilization of timestamp-predictive and instance-contrastive tasks for disentangled representation learning, with the former optimizing timestamp-level embeddings with predictive loss, and the latter optimizing instance-level embeddings with contrastive loss; and (iii) avoidance of augmentation methods to eliminate inductive biases, such as transformation-invariance from cropping and masking. Comprehensive experiments on 6 time-series forecasting datasets and 5 time-series classification datasets have shown that TimeDRL consistently surpasses existing representation learning approaches, achieving an average improvement of forecasting by 58.02% in MSE and classification by 1.48% in accuracy. Furthermore, extensive ablation studies confirmed the relative contribution of each component in TimeDRL's architecture, and semi-supervised learning evaluations demonstrated its effectiveness in real-world scenarios, even with limited labeled data. The code is available at https://github.com/blacksnail789521/TimeDRL.

cs.LG↗

arXiv2024

PlayBest: Professional Basketball Player Behavior Synthesis via Planning with Diffusion

Dynamically planning in complex systems has been explored to improve decision-making in various domains. Professional basketball serves as a compelling example of a dynamic spatio-temporal game, encompassing context-dependent decision-making. However, processing the diverse on-court signals and navigating the vast space of potential actions and outcomes make it difficult for existing approaches to swiftly identify optimal strategies in response to evolving circumstances. In this study, we formulate the sequential decision-making process as a conditional trajectory generation process. Based on the formulation, we introduce PlayBest (PLAYer BEhavior SynThesis), a method to improve player decision-making. We extend the diffusion probabilistic model to learn challenging environmental dynamics from historical National Basketball Association (NBA) player motion tracking data. To incorporate data-driven strategies, an auxiliary value function is trained with corresponding rewards. To accomplish reward-guided trajectory generation, we condition the diffusion model on the value function via classifier-guided sampling. We validate the effectiveness of PlayBest through simulation studies, contrasting the generated trajectories with those employed by professional basketball teams. Our results reveal that the model excels at generating reasonable basketball trajectories that produce efficient plays. Moreover, the synthesized play strategies exhibit an alignment with professional tactics, highlighting the model's capacity to capture the intricate dynamics of basketball games.

cs.AI↗

arXiv2024

Benchmarking Stroke Forecasting with Stroke-Level Badminton Dataset

In recent years, badminton analytics has drawn attention due to the advancement of artificial intelligence and the efficiency of data collection. While there is a line of effective applications to improve and investigate player performance, there are only a few public badminton datasets that can be used by researchers outside the badminton domain. Existing badminton singles datasets focus on specific matchups; however, they cannot provide comprehensive studies on different players and various matchups. In this paper, we provide a badminton singles dataset, ShuttleSet22, which is collected from high-ranking matches in 2022. ShuttleSet22 consists of 30,172 strokes in 2,888 rallies in the training set, 1,400 strokes in 450 rallies in the validation set, and 2,040 strokes in 654 rallies in the testing set, with detailed stroke-level metadata within a rally. To benchmark existing work with ShuttleSet22, we hold a challenge, Track 2: Forecasting Future Turn-Based Strokes in Badminton Rallies, at CoachAI Badminton Challenge @ IJCAI 2023, to encourage researchers to tackle this real-world problem through innovative approaches and to summarize insights between the state-of-the-art baseline and improved techniques, exchanging inspiring ideas. The baseline codes and the dataset are made available at https://github.com/wywyWang/CoachAI-Projects/tree/main/CoachAI-Challenge-IJCAI2023.

cs.AI↗