arXiv subjects

Kim-Hui Yap

Publications and source records attributed to Kim-Hui Yap.

At least 19 recordsLinked to original sources

arXiv2026

SWE-Review: Closing the Loop on Issue Resolution with Agentic Code Review

Coding agents increasingly generate pull requests (PRs) for real-world software issues, yet one-shot PR generation remains open-loop: the PR is proposed without systematic review, diagnosis, or revision. We introduce \textbf{SWE-Review}, a framework for closing this loop with agentic code review. Given an issue and an AI-generated PR, a reviewer agent explores the repository, decides whether the PR should be accepted, and provides structured feedback for revision. We evaluate this setting with our proposed \textbf{SWE-Review-Bench} to measure both review correctness and downstream revision usefulness. We further curate \textbf{SWE-Review-Traj} dataset to study broader applications of agentic review and fill the data-scarcity gap for open reviewer training. Experiments show that agentic review continuously improves PRs through a generate-review-revise loop, outperforms single-turn fixed-context review in both decision accuracy and resolve rate after revision, transfers beyond review to improve issue-resolution models, and enables effective and efficient test-time scaling. These results position agentic code review as a practical mechanism for moving AI coding agents from one-shot PR generation toward closed-loop issue resolution.

cs.SE↗

arXiv2026

S-Agent: Spatial Tool-Use Elicits Reasoning for Spatial Intelligence

Real-world spatial intelligence requires reasoning over a continuous and evolving 3D world, yet existing VLMs and tool-augmented agents largely remain tied to static, stateless inference from isolated visual observations. We introduce \textbf{\textsc{S-Agent}}, a spatial tool-use agentic paradigm for understanding and reasoning over continuous multi-view images and videos. By formulating spatial reasoning as spatio-temporal evidence accumulation rather than isolated frame-level prediction, \textsc{S-Agent} reshapes spatial perception into scene-centric understanding beyond frame-centric recognition. Specifically, \textsc{S-Agent} casts the VLM as a semantic planner that decides what evidence is needed, while a hierarchy of spatial tools and experts grounds objects in 2D, lifts them into 3D geometric evidence, and aggregates this evidence into high-level spatial knowledge (\textit{e.g.}, counting, measurement, orientation, and relative position). Additionally, a temporal memory mechanism, including Scene Memory for maintaining the evolving scene state and Agent Memory for accumulating reasoning context, enables evidence integration across frames and reasoning steps. Comprehensive experiments on multi-view and video spatial reasoning benchmarks show that \textsc{S-Agent} consistently improves both open-source and closed-source VLMs in a training-free manner. Beyond inference-time augmentation, supervised fine-tuning (SFT) on \textsc{S-Agent}-generated spatial trajectories \textsc{S-300K} yields \textsc{S-Agent-8B}, a compact spatial agent that significantly surpasses similar-scale baselines (e.g., Qwen3-VL-8B) and performs comparably to advanced closed-source models (e.g., GPT-5.4 and Gemini 3).

cs.CV↗

arXiv2026

Demystifying Data Organization for Enhanced LLM Training

Large Language Models (LLMs) have revolutionized various fields, yet their training efficiency is heavily reliant on effective data curation. While data selection has been widely studied, the strategic data organization for enhanced training remains an underexplored area, particularly since current LLMs are often trained for only one or a few epochs. This paper systematically explores the influence of data organization on LLM training by reusing pre-computed sample-level scores originally generated for data efficiency, thereby incurring minimal additional computational overhead. We identify and formalize four key guidelines for optimizing data organization: Boundary Sharpening, Cyclic Scheduling, Curriculum Continuity, and Local Diversity. Guided by them, we introduce two novel data ordering methods termed STR and SAW. Extensive experiments across different model scales and data sizes, encompassing both pre-training and SFT stages, validate the effectiveness of our summarized guidelines. They also demonstrate the robustness of our proposed data ordering methods in enhancing the stability and performance of LLM training. Github Link: https://github.com/microsoft/data-efficacy/

cs.AI↗

arXiv2026

NTIRE 2026 Challenge on Bitstream-Corrupted Video Restoration: Methods and Results

This paper reports on the NTIRE 2026 Challenge on Bitstream-Corrupted Video Restoration (BSCVR). The challenge aims to advance research on recovering visually coherent videos from corrupted bitstreams, whose decoding often produces severe spatial-temporal artifacts and content distortion. Built upon recent progress in bitstream-corrupted video recovery, the challenge provides a common benchmark for evaluating restoration methods under realistic corruption settings. We describe the dataset, evaluation protocol, and participating methods, and summarize the final results and main technical trends. The challenge highlights the difficulty of this emerging task and provides useful insights for future research on robust video restoration under practical bitstream corruption.

cs.CV↗

arXiv2026

Mixture-of-Modality-Experts with Holistic Token Learning for Fine-Grained Multimodal Visual Analytics in Driver Action Recognition

Robust multimodal visual analytics remains challenging when heterogeneous modalities provide complementary but input-dependent evidence for decision-making.Existing multimodal learning methods mainly rely on fixed fusion modules or predefined cross-modal interactions, which are often insufficient to adapt to changing modality reliability and to capture fine-grained action cues. To address this issue, we propose a Mixture-of-Modality-Experts (MoME) framework with a Holistic Token Learning (HTL) strategy. MoME enables adaptive collaboration among modality-specific experts, while HTL improves both intra-expert refinement and inter-expert knowledge transfer through class tokens and spatio-temporal tokens. In this way, our method forms a knowledge-centric multimodal learning framework that improves expert specialization while reducing ambiguity in multimodal fusion.We validate the proposed framework on driver action recognition as a representative multimodal understanding taskThe experimental results on the public benchmark show that the proposed MoME framework and the HTL strategy jointly outperform representative single-modal and multimodal baselines. Additional ablation, validation, and visualization results further verify that the proposed HTL strategy improves subtle multimodal understanding and offers better interpretability.

cs.CV↗

arXiv2026

Accelerating Diffusion-based Video Editing via Heterogeneous Caching: Beyond Full Computing at Sampled Denoising Timestep

Diffusion-based video editing has emerged as an important paradigm for high-quality and flexible content generation. However, despite their generality and strong modeling capacity, Diffusion Transformers (DiT) remain computationally expensive due to the iterative denoising process, posing challenges for practical deployment. Existing video diffusion acceleration methods primarily exploit denoising timestep-level feature reuse, which mitigates the redundancy in denoising process, but overlooks the architectural redundancy within the DiT that many attention operations over spatio-temporal tokens are redundantly executed, offering little to no incremental contribution to the model output. This work introduces HetCache, a training-free diffusion acceleration framework designed to exploit the inherent heterogeneity in diffusion-based masked video-to-video (MV2V) generation and editing. Instead of uniformly reuse or randomly sampling tokens, HetCache assesses the contextual relevance and interaction strength among various types of tokens in designated computing steps. Guided by spatial priors, it divides the spatial-temporal tokens in DiT model into context and generative tokens, and selectively caches the context tokens that exhibit the strongest correlation and most representative semantics with generative ones. This strategy reduces redundant attention operations while maintaining editing consistency and fidelity. Experiments show that HetCache achieves a noticeable acceleration, including a 2.67$\times$ latency speedup and FLOPs reduction over commonly used foundation models, with negligible degradation in editing quality.

cs.CV↗

arXiv2026

MuseAgent-1: Interactive Grounded Multimodal Understanding of Music Scores and Performance Audio

Despite recent advances in multimodal large language models (MLLMs), their ability to understand and interact with music remains limited. Music understanding requires grounded reasoning over symbolic scores and expressive performance audio, which general-purpose MLLMs often fail to handle due to insufficient perceptual grounding. We introduce MuseAgent, a music-centric multimodal agent that augments language models with structured symbolic representations derived from sheet music images and performance audio. By integrating optical music recognition and automatic music transcription modules, MuseAgent enables multi-step reasoning and interaction over fine-grained musical content. To systematically evaluate music understanding capabilities, we further propose MuseBench, a benchmark covering music theory reasoning, score interpretation, and performance-level analysis across text, image, and audio modalities. Experiments show that existing MLLMs perform poorly on these tasks, while MuseAgent achieves substantial improvements, highlighting the importance of structured multimodal grounding for interactive music understanding.

cs.MM↗

arXiv2026

DAOS: A Multimodal In-cabin Behavior Monitoring with Driver Action-Object Synergy Dataset

In driver activity monitoring, movements are mostly limited to the upper body, which makes many actions look similar. To tell these actions apart, human often rely on the objects the driver is using, such as holding a phone compared with gripping the steering wheel. However, most existing driver-monitoring datasets lack accurate object-location annotations or do not link objects to their associated actions, leaving a critical gap for reliable action recognition. To address this, we introduce the Driver Action with Object Synergy (DAOS) dataset, comprising 9,787 video clips annotated with 36 fine-grained driver actions and 15 object classes, totaling more than 2.5 million corresponding object instances. DAOS offers multi-modal, multi-view data (RGB, IR, and depth) from front, face, left, and right perspectives. Although DAOS captures a wide range of cabin objects, only a few are directly relevant to each action for prediction, so focusing on task-specific human-object relations is essential. To tackle this challenge, we propose the Action-Object-Relation Network (AOR-Net). AOR-Net comprehends complex driver actions through multi-level reasoning and a chain-of-action prompting mechanism that models the logical relationships among actions, objects, and their relations. Additionally, the Mixture of Thoughts module is introduced to dynamically select essential knowledge at each stage, enhancing robustness in object-rich and object-scarce conditions. Extensive experiments demonstrate that our model outperforms other state-of-the-art methods on various datasets.

cs.CV↗

arXiv2025

Building Egocentric Procedural AI Assistant: Methods, Benchmarks, and Challenges

Driven by recent advances in vision-language models (VLMs) and egocentric perception research, the emerging topic of an egocentric procedural AI assistant (EgoProceAssist) is introduced to step-by-step support daily procedural tasks in a first-person view. In this paper, we start by identifying three core tasks in EgoProceAssist: egocentric procedural error detection, egocentric procedural learning, and egocentric procedural question answering, then introduce two enabling dimensions: real-time and streaming video understanding, and proactive interaction in procedural contexts. We define these tasks within a new taxonomy as the EgoProceAssist's essential functions and illustrate how they can be deployed in real-world scenarios for daily activity assistants. Specifically, our work encompasses a comprehensive review of current techniques, relevant datasets, and evaluation metrics across these five core areas. To clarify the gap between the proposed EgoProceAssist and existing VLM-based assistants, we conduct novel experiments to provide a comprehensive evaluation of representative VLM-based methods. Through these findings and our technical analysis, we discuss the challenges ahead and suggest future research directions. Furthermore, an exhaustive list of this study is publicly available in an active repository that continuously collects the latest work: https://github.com/z1oong/Building-Egocentric-Procedural-AI-Assistant.

cs.CV↗

arXiv2025

Towards Blind Bitstream-corrupted Video Recovery via a Visual Foundation Model-driven Framework

Video signals are vulnerable in multimedia communication and storage systems, as even slight bitstream-domain corruption can lead to significant pixel-domain degradation. To recover faithful spatio-temporal content from corrupted inputs, bitstream-corrupted video recovery has recently emerged as a challenging and understudied task. However, existing methods require time-consuming and labor-intensive annotation of corrupted regions for each corrupted video frame, resulting in a large workload in practice. In addition, high-quality recovery remains difficult as part of the local residual information in corrupted frames may mislead feature completion and successive content recovery. In this paper, we propose the first blind bitstream-corrupted video recovery framework that integrates visual foundation models with a recovery model, which is adapted to different types of corruption and bitstream-level prompts. Within the framework, the proposed Detect Any Corruption (DAC) model leverages the rich priors of the visual foundation model while incorporating bitstream and corruption knowledge to enhance corruption localization and blind recovery. Additionally, we introduce a novel Corruption-aware Feature Completion (CFC) module, which adaptively processes residual contributions based on high-level corruption understanding. With VFM-guided hierarchical feature augmentation and high-level coordination in a mixture-of-residual-experts (MoRE) structure, our method suppresses artifacts and enhances informative residuals. Comprehensive evaluations show that the proposed method achieves outstanding performance in bitstream-corrupted video recovery without requiring a manually labeled mask sequence. The demonstrated effectiveness will help to realize improved user experience, wider application scenarios, and more reliable multimedia communication and storage systems.

eess.IV↗

arXiv2025

A Structure-aware and Motion-adaptive Framework for 3D Human Pose Estimation with Mamba

Recent Mamba-based methods for the pose-lifting task tend to model joint dependencies by 2D-to-1D mapping with diverse scanning strategies. Though effective, they struggle to model intricate joint connections and uniformly process all joint motion trajectories while neglecting the intrinsic differences across motion characteristics. In this work, we propose a structure-aware and motion-adaptive framework to capture spatial joint topology along with diverse motion dynamics independently, named as SAMA. Specifically, SAMA consists of a Structure-aware State Integrator (SSI) and a Motion-adaptive State Modulator (MSM). The Structure-aware State Integrator is tasked with leveraging dynamic joint relationships to fuse information at both the joint feature and state levels in the state space, based on pose topology rather than sequential state transitions. The Motion-adaptive State Modulator is responsible for joint-specific motion characteristics recognition, thus applying tailored adjustments to diverse motion patterns across different joints. Through the above key modules, our algorithm enables structure-aware and motion-adaptive pose lifting. Extensive experiments across multiple benchmarks demonstrate that our algorithm achieves advanced results with fewer computational costs.

cs.CV↗

arXiv2025

PromptSR: Cascade Prompting for Lightweight Image Super-Resolution

Although the lightweight Vision Transformer has significantly advanced image super-resolution (SR), it faces the inherent challenge of a limited receptive field due to the window-based self-attention modeling. The quadratic computational complexity relative to window size restricts its ability to use a large window size for expanding the receptive field while maintaining low computational costs. To address this challenge, we propose PromptSR, a novel prompt-empowered lightweight image SR method. The core component is the proposed cascade prompting block (CPB), which enhances global information access and local refinement via three cascaded prompting layers: a global anchor prompting layer (GAPL) and two local prompting layers (LPLs). The GAPL leverages downscaled features as anchors to construct low-dimensional anchor prompts (APs) through cross-scale attention, significantly reducing computational costs. These APs, with enhanced global perception, are then used to provide global prompts, efficiently facilitating long-range token connections. The two LPLs subsequently combine category-based self-attention and window-based self-attention to refine the representation in a coarse-to-fine manner. They leverage attention maps from the GAPL as additional global prompts, enabling them to perceive features globally at different granularities for adaptive local refinement. In this way, the proposed CPB effectively combines global priors and local details, significantly enlarging the receptive field while maintaining the low computational costs of our PromptSR. The experimental results demonstrate the superiority of our method, which outperforms state-of-the-art lightweight SR methods in quantitative, qualitative, and complexity evaluations. Our code will be released at https://github.com/wenyang001/PromptSR.

cs.CV↗

arXiv2025

SSH-Net: A Self-Supervised and Hybrid Network for Noisy Image Watermark Removal

Visible watermark removal is challenging due to its inherent complexities and the noise carried within images. Existing methods primarily rely on supervised learning approaches that require paired datasets of watermarked and watermark-free images, which are often impractical to obtain in real-world scenarios. To address this challenge, we propose SSH-Net, a Self-Supervised and Hybrid Network specifically designed for noisy image watermark removal. SSH-Net synthesizes reference watermark-free images using the watermark distribution in a self-supervised manner and adopts a dual-network design to address the task. The upper network, focused on the simpler task of noise removal, employs a lightweight CNN-based architecture, while the lower network, designed to handle the more complex task of simultaneously removing watermarks and noise, incorporates Transformer blocks to model long-range dependencies and capture intricate image features. To enhance the model's effectiveness, a shared CNN-based feature encoder is introduced before dual networks to extract common features that both networks can leverage. Our code will be available at https://github.com/wenyang001/SSH-Net.

cs.MM↗

arXiv2024

ByteNet: Rethinking Multimedia File Fragment Classification through Visual Perspectives

Multimedia file fragment classification (MFFC) aims to identify file fragment types, e.g., image/video, audio, and text without system metadata. It is of vital importance in multimedia storage and communication. Existing MFFC methods typically treat fragments as 1D byte sequences and emphasize the relations between separate bytes (interbytes) for classification. However, the more informative relations inside bytes (intrabytes) are overlooked and seldom investigated. By looking inside bytes, the bit-level details of file fragments can be accessed, enabling a more accurate classification. Motivated by this, we first propose Byte2Image, a novel visual representation model that incorporates previously overlooked intrabyte information into file fragments and reinterprets these fragments as 2D grayscale images. This model involves a sliding byte window to reveal the intrabyte information and a rowwise stacking of intrabyte ngrams for embedding fragments into a 2D space. Thus, complex interbyte and intrabyte correlations can be mined simultaneously using powerful vision networks. Additionally, we propose an end-to-end dual-branch network ByteNet to enhance robust correlation mining and feature representation. ByteNet makes full use of the raw 1D byte sequence and the converted 2D image through a shallow byte branch feature extraction (BBFE) and a deep image branch feature extraction (IBFE) network. In particular, the BBFE, composed of a single fully-connected layer, adaptively recognizes the co-occurrence of several some specific bytes within the raw byte sequence, while the IBFE, built on a vision Transformer, effectively mines the complex interbyte and intrabyte correlations from the converted image. Experiments on the two representative benchmarks, including 14 cases, validate that our proposed method outperforms state-of-the-art approaches on different cases by up to 12.2%.

cs.CV↗

arXiv2024

CL-HOI: Cross-Level Human-Object Interaction Distillation from Vision Large Language Models

Human-object interaction (HOI) detection has seen advancements with Vision Language Models (VLMs), but these methods often depend on extensive manual annotations. Vision Large Language Models (VLLMs) can inherently recognize and reason about interactions at the image level but are computationally heavy and not designed for instance-level HOI detection. To overcome these limitations, we propose a Cross-Level HOI distillation (CL-HOI) framework, which distills instance-level HOIs from VLLMs image-level understanding without the need for manual annotations. Our approach involves two stages: context distillation, where a Visual Linguistic Translator (VLT) converts visual information into linguistic form, and interaction distillation, where an Interaction Cognition Network (ICN) reasons about spatial, visual, and context relations. We design contrastive distillation losses to transfer image-level context and interaction knowledge from the teacher to the student model, enabling instance-level HOI detection. Evaluations on HICO-DET and V-COCO datasets demonstrate that our CL-HOI surpasses existing weakly supervised methods and VLLM supervised methods, showing its efficacy in detecting HOIs without manual labels.

cs.CV↗

arXiv2024

Open World Object Detection: A Survey

Exploring new knowledge is a fundamental human ability that can be mirrored in the development of deep neural networks, especially in the field of object detection. Open world object detection (OWOD) is an emerging area of research that adapts this principle to explore new knowledge. It focuses on recognizing and learning from objects absent from initial training sets, thereby incrementally expanding its knowledge base when new class labels are introduced. This survey paper offers a thorough review of the OWOD domain, covering essential aspects, including problem definitions, benchmark datasets, source codes, evaluation metrics, and a comparative study of existing methods. Additionally, we investigate related areas like open set recognition (OSR) and incremental learning (IL), underlining their relevance to OWOD. Finally, the paper concludes by addressing the limitations and challenges faced by current OWOD algorithms and proposes directions for future research. To our knowledge, this is the first comprehensive survey of the emerging OWOD field with over one hundred references, marking a significant step forward for object detection technology. A comprehensive source code and benchmarks are archived and concluded at https://github.com/ArminLee/OWOD Review.

cs.CV↗

arXiv2024

Empowering Large Language Model for Continual Video Question Answering with Collaborative Prompting

In recent years, the rapid increase in online video content has underscored the limitations of static Video Question Answering (VideoQA) models trained on fixed datasets, as they struggle to adapt to new questions or tasks posed by newly available content. In this paper, we explore the novel challenge of VideoQA within a continual learning framework, and empirically identify a critical issue: fine-tuning a large language model (LLM) for a sequence of tasks often results in catastrophic forgetting. To address this, we propose Collaborative Prompting (ColPro), which integrates specific question constraint prompting, knowledge acquisition prompting, and visual temporal awareness prompting. These prompts aim to capture textual question context, visual content, and video temporal dynamics in VideoQA, a perspective underexplored in prior research. Experimental results on the NExT-QA and DramaQA datasets show that ColPro achieves superior performance compared to existing approaches, achieving 55.14\% accuracy on NExT-QA and 71.24\% accuracy on DramaQA, highlighting its practical relevance and effectiveness.

cs.CV↗

arXiv2024

MultiFuser: Multimodal Fusion Transformer for Enhanced Driver Action Recognition

Driver action recognition, aiming to accurately identify drivers' behaviours, is crucial for enhancing driver-vehicle interactions and ensuring driving safety. Unlike general action recognition, drivers' environments are often challenging, being gloomy and dark, and with the development of sensors, various cameras such as IR and depth cameras have emerged for analyzing drivers' behaviors. Therefore, in this paper, we propose a novel multimodal fusion transformer, named MultiFuser, which identifies cross-modal interrelations and interactions among multimodal car cabin videos and adaptively integrates different modalities for improved representations. Specifically, MultiFuser comprises layers of Bi-decomposed Modules to model spatiotemporal features, with a modality synthesizer for multimodal features integration. Each Bi-decomposed Module includes a Modal Expertise ViT block for extracting modality-specific features and a Patch-wise Adaptive Fusion block for efficient cross-modal fusion. Extensive experiments are conducted on Drive&Act dataset and the results demonstrate the efficacy of our proposed approach.

cs.CV↗