arXiv subjects

Lingyun Sun

Publications and source records attributed to Lingyun Sun.

At least 19 recordsLinked to original sources

arXiv2026

MoWorld: A Flash World Model

The future of World Models depends not only on scaling model capability, but also on scaling practicality and inference efficiency. High-frame-rate inference enables responsive perception, planning, and control in real-world autonomous systems. To this end, we present MoWorld, a cost-effective yet high-performance Flash World Model with an end-to-end framework spanning data generation, pre-training, distillation, and efficient inference, enabling up to 50 FPS real-time interaction with cinematic visual quality without the need of high-end GPUs. To enable large-scale real-world deployment, MoWorld jointly optimizes model capability and cost throughout the entire development pipeline. Specifically, unlike existing approaches that primarily rely on large-scale video corpora, MoWorld is built upon a scalable 3D-native data engine accumulated from our large-scale 3D vision and generative modeling pipeline, enabling the efficient construction of geometrically consistent training data across diverse real-world and synthetic environments. Based on this foundation, a curriculum cross-frame pre-training strategy for stable and scalable World Model learning, an efficient denoising-step distillation algorithm to reduce diffusion training cost, and a mixed-precision parallel inference framework for low-cost real-time deployment. MoWorld is the first real-time interactive World Model built on the Neural Processing Unit (NPU) and can achieves up to 50 FPS in such the devices, enabling practical and efficient deployment at scale. Comprehensive evaluations demonstrate that MoWorld achieves leading performance; notably, its average inference cost is only 30\%-50\% of that of existing World Models, providing a practical foundation for large-scale real-world applications of World Models. We also demonstrate diverse applications of MoWorld.

cs.CV↗

arXiv2026

Mean Flow Distillation: Robust and Stable Distillation for Flow Matching Models

Flow Matching models have demonstrated strong performance across a wide range of generative tasks. However, their reliance on ODE-based iterative sampling incurs substantial computational overhead in inference, which limits their applicability in real-time scenes. While distillation is a promising solution, existing approaches largely borrow from diffusion-based score matching, often failing to exploit the intrinsic geometric structure of flows and suffering from training instability, high variance, and degraded generation quality. In this paper, we propose Mean Flow Distillation (MFD), a novel distillation framework tailored for flow matching models. We theoretically demonstrate that MFD acts as a temporal low-pass filter, effectively suppressing the high-frequency optimization noise inherent in variational score distillation (VSD) while ensuring global trajectory consistency. We further prove the Mean Flow Matching Theorem, establishing that matching expected average velocities is sufficient for strict distribution alignment. Empirically, on challenging tasks of high-dimensional manifolds including 4D occupancy forecasting and text-to-image generation, MFD achieves state-of-the-art performance, enabling high-fidelity single-step generation.

cs.CV↗

arXiv2026

4DVGGT-D: 4D Visual Geometry Transformer with Improved Dynamic Depth Estimation

Reconstructing dynamic 4D scenes from monocular videos is a fundamental yet challenging task. While recent 3D foundation models provide strong geometric priors, their performance significantly degrades in dynamic environments. This degradation stems from a fundamental tension: the inherent coupling of camera ego-motion and object motion within global attention mechanisms. In this paper, we propose a novel, training-free progressive decoupling framework that disentangles dynamics from statics in a principled, coarse-to-fine manner. Our core insight is to resolve the tension by first stabilizing the camera pose, followed by geometric refinement. Specifically, our approach consists of three synergistic components: (1) a Dynamic-Mask-Guided Pose Decoupling module that isolates pose estimation from dynamic interference, yielding a stable motion-free reference frame; (2) a Topological Subspace Surgery mechanism that orthogonally decomposes the depth manifold, safely preserving dynamic objects while injecting refined, mask-aware geometry into static regions; and (3) an Information-Theoretic Confidence-Aware Fusion strategy that formulates depth integration as a heteroscedastic Bayesian inference problem, adaptively blending multi-pass predictions via inverse-variance weighting. Extensive experiments on standard 4D reconstruction benchmarks demonstrate that our method achieves consistent and substantial improvements across principal point-cloud metrics. Notably, our approach shows competitive performance in robust 4D scene reconstruction without requiring fine-tuning, suggesting the potential of mathematically grounded dynamic-static disentanglement.

cs.CV↗

arXiv2026

StreamCacheVGGT: Streaming Visual Geometry Transformers with Robust Scoring and Hybrid Cache Compression

Reconstructing dense 3D geometry from continuous video streams requires stable inference under a constant memory budget. Existing $O(1)$ frameworks primarily rely on a ``pure eviction'' paradigm, which suffers from significant information destruction due to binary token deletion and evaluation noise from localized, single-layer scoring. To address these bottlenecks, we propose StreamCacheVGGT, a training-free framework that reimagines cache management through two synergistic modules: Cross-Layer Consistency-Enhanced Scoring (CLCES) and Hybrid Cache Compression (HCC). CLCES mitigates activation noise by tracking token importance trajectories across the Transformer hierarchy, employing order-statistical analysis to identify sustained geometric salience. Leveraging these robust scores, HCC transcends simple eviction by introducing a three-tier triage strategy that merges moderately important tokens into retained anchors via nearest-neighbor assignment on the key-vector manifold. This approach preserves essential geometric context that would otherwise be lost. Extensive evaluations on five benchmarks (7-Scenes, NRGBD, ETH3D, Bonn, and KITTI) demonstrate that StreamCacheVGGT sets a new state-of-the-art, delivering superior reconstruction accuracy and long-term stability while strictly adhering to constant-cost constraints.

cs.CV↗

arXiv2026

Robust 4D Visual Geometry Transformer with Uncertainty-Aware Priors

Reconstructing dynamic 4D scenes is an important yet challenging task. While 3D foundation models like VGGT excel in static settings, they often struggle with dynamic sequences where motion causes significant geometric ambiguity. To address this, we present a framework designed to disentangle dynamic and static components by modeling uncertainty across different stages of the reconstruction process. Our approach introduces three synergistic mechanisms: (1) Entropy-Guided Subspace Projection, which leverages information-theoretic weighting to adaptively aggregate multi-head attention distributions, effectively isolating dynamic motion cues from semantic noise; (2) Local-Consistency Driven Geometry Purification, which enforces spatial continuity via radius-based neighborhood constraints to eliminate structural outliers; and (3) Uncertainty-Aware Cross-View Consistency, which formulates multi-view projection refinement as a heteroscedastic maximum likelihood estimation problem, utilizing depth confidence as a probabilistic weight. Experiments on dynamic benchmarks show that our approach outperforms current state-of-the-art methods, reducing Mean Accuracy error by 13.43\% and improving segmentation F-measure by 10.49\%. Our framework maintains the efficiency of feed-forward inference and requires no task-specific fine-tuning or per-scene optimization.

cs.CV↗

arXiv2026

Seeing Eye to Eye: Enabling Cognitive Alignment Through Shared First-Person Perspective in Human-AI Collaboration

Despite advances in multimodal AI, current vision-based assistants often remain inefficient in collaborative tasks. We identify two key gulfs: a communication gulf, where users must translate rich parallel intentions into verbal commands due to the channel mismatch , and an understanding gulf, where AI struggles to interpret subtle embodied cues. To address these, we propose Eye2Eye, a framework that leverages first-person perspective as a channel for human-AI cognitive alignment. It integrates three components: (1) joint attention coordination for fluid focus alignment, (2) revisable memory to maintain evolving common ground, and (3) reflective feedback allowing users to clarify and refine AI's understanding. We implement this framework in an AR prototype and evaluate it through a user study and a post-hoc pipeline evaluation. Results show that Eye2Eye significantly reduces task completion time and interaction load while increasing trust, demonstrating its components work in concert to improve collaboration.

cs.HC↗

arXiv2026

Beyond Binary Preference: Aligning Diffusion Models to Fine-grained Criteria by Decoupling Attributes

Post-training alignment of diffusion models relies on simplified signals, such as scalar rewards or binary preferences. This limits alignment with complex human expertise, which is hierarchical and fine-grained. To address this, we first construct a hierarchical, fine-grained evaluation criteria with domain experts, which decomposes image quality into multiple positive and negative attributes organized in a tree structure. Building on this, we propose a two-stage alignment framework. First, we inject domain knowledge to an auxiliary diffusion model via Supervised Fine-Tuning. Second, we introduce Complex Preference Optimization (CPO) that extends DPO to align the target diffusion to our non-binary, hierarchical criteria. Specifically, we reformulate the alignment problem to simultaneously maximize the probability of positive attributes while minimizing the probability of negative attributes with the auxiliary diffusion. We instantiate our approach in the domain of painting generation and conduct CPO training with an annotated dataset of painting with fine-grained attributes based on our criteria. Extensive experiments demonstrate that CPO significantly enhances generation quality and alignment with expertise, opening new avenues for fine-grained criteria alignment.

cs.CV↗

arXiv2025

Visionary Co-Driver: Enhancing Driver Perception of Potential Risks with LLM and HUD

Drivers' perception of risky situations has always been a challenge in driving. Existing risk-detection methods excel at identifying collisions but face challenges in assessing the behavior of road users in non-collision situations. This paper introduces Visionary Co-Driver, a system that leverages large language models to identify non-collision roadside risks and alert drivers based on their eye movements. Specifically, the system combines video processing algorithms and LLMs to identify potentially risky road users. These risks are dynamically indicated on an adaptive heads-up display interface to enhance drivers' attention. A user study with 41 drivers confirms that Visionary Co-Driver improves drivers' risk perception and supports their recognition of roadside risks.

cs.HC↗

arXiv2025

Learning to Hear by Seeing: It's Time for Vision Language Models to Understand Artistic Emotion from Sight and Sound

Emotion understanding is critical for making Large Language Models (LLMs) more general, reliable, and aligned with humans. Art conveys emotion through the joint design of visual and auditory elements, yet most prior work is human-centered or single-modality, overlooking the emotion intentionally expressed by the artwork. Meanwhile, current Audio-Visual Language Models (AVLMs) typically require large-scale audio pretraining to endow Visual Language Models (VLMs) with hearing, which limits scalability. We present Vision Anchored Audio-Visual Emotion LLM (VAEmotionLLM), a two-stage framework that teaches a VLM to hear by seeing with limited audio pretraining and to understand emotion across modalities. In Stage 1, Vision-Guided Audio Alignment (VG-Align) distills the frozen visual pathway into a new audio pathway by aligning next-token distributions of the shared LLM on synchronized audio-video clips, enabling hearing without a large audio dataset. In Stage 2, a lightweight Cross-Modal Emotion Adapter (EmoAdapter), composed of the Emotion Enhancer and the Emotion Supervisor, injects emotion-sensitive residuals and applies emotion supervision to enhance cross-modal emotion understanding. We also construct ArtEmoBenchmark, an art-centric emotion benchmark that evaluates content and emotion understanding under audio-only, visual-only, and audio-visual inputs. VAEmotionLLM achieves state-of-the-art results on ArtEmoBenchmark, outperforming audio-only, visual-only, and audio-visual baselines. Ablations show that the proposed components are complementary.

cs.CV↗

arXiv2025

SCENIC: A Location-based System to Foster Cognitive Development in Children During Car Rides

Car-riding is common for children in modern life. Given the repetitive nature of daily commutes, they often feel bored and turn to electronic devices for entertainment. Meanwhile, the rich and dynamic scenery outside the car naturally attracts children's curiosity and offers valuable resources for cognitive development. Our formative study reveals that parents' support during car rides is often fleeting, as accompanying adults may struggle to consistently guide children's exploration. To address this, we propose SCENIC, an interactive system that helps children aged 6 to 11 better perceive the external environment using location-based cognitive development strategies. SCENIC builds upon experiential approaches used by parents, resulting in six strategies embedded into the system. To improve engagement during routine rides, SCENIC also incorporates dynamic point-of-interest selection and journey gallery generation. We evaluated the generated content (N=21) and conducted an in-situ user study with seven families and ten children. Results suggest that SCENIC enhances the car-riding experience and helps children better connect with their surroundings.

cs.HC↗

arXiv2025

Hand by Hand: LLM Driving EMS Assistant for Operational Skill Learning

Operational skill learning, inherently physical and reliant on hands-on practice and kinesthetic feedback, has yet to be effectively replicated in large language model (LLM)-supported training. Current LLM training assistants primarily generate customized textual feedback, neglecting the crucial kinesthetic modality. This gap derives from the textual and uncertain nature of LLMs, compounded by concerns on user acceptance of LLM driven body control. To bridge this gap and realize the potential of collaborative human-LLM action, this work explores human experience of LLM driven kinesthetic assistance. Specifically, we introduced an "Align-Analyze-Adjust" strategy and developed FlightAxis, a tool that integrates LLM with Electrical Muscle Stimulation (EMS) for flight skill acquisition, a representative operational skill domain. FlightAxis learns flight skills from manuals and guides forearm movements during simulated flight tasks. Our results demonstrate high user acceptance of LLM-mediated body control and significantly reduced task completion times. Crucially, trainees reported that this kinesthetic assistance enhanced their awareness of operation flaws and fostered increased engagement in the training process, rather than relieving perceived load. This work demonstrated the potential of kinesthetic LLM training in operational skill acquisition.

cs.HC↗

arXiv2025

Driver Assistant: Persuading Drivers to Adjust Secondary Tasks Using Large Language Models

Level 3 automated driving systems allows drivers to engage in secondary tasks while diminishing their perception of risk. In the event of an emergency necessitating driver intervention, the system will alert the driver with a limited window for reaction and imposing a substantial cognitive burden. To address this challenge, this study employs a Large Language Model (LLM) to assist drivers in maintaining an appropriate attention on road conditions through a "humanized" persuasive advice. Our tool leverages the road conditions encountered by Level 3 systems as triggers, proactively steering driver behavior via both visual and auditory routes. Empirical study indicates that our tool is effective in sustaining driver attention with reduced cognitive load and coordinating secondary tasks with takeover behavior. Our work provides insights into the potential of using LLMs to support drivers during multi-task automated driving.

cs.HC↗

arXiv2025

Controllable Video-to-Music Generation with Multiple Time-Varying Conditions

Music enhances video narratives and emotions, driving demand for automatic video-to-music (V2M) generation. However, existing V2M methods relying solely on visual features or supplementary textual inputs generate music in a black-box manner, often failing to meet user expectations. To address this challenge, we propose a novel multi-condition guided V2M generation framework that incorporates multiple time-varying conditions for enhanced control over music generation. Our method uses a two-stage training strategy that enables learning of V2M fundamentals and audiovisual temporal synchronization while meeting users' needs for multi-condition control. In the first stage, we introduce a fine-grained feature selection module and a progressive temporal alignment attention mechanism to ensure flexible feature alignment. For the second stage, we develop a dynamic conditional fusion module and a control-guided decoder module to integrate multiple conditions and accurately guide the music composition process. Extensive experiments demonstrate that our method outperforms existing V2M pipelines in both subjective and objective evaluations, significantly enhancing control and alignment with user expectations.

cs.MM↗

arXiv2025

Inversion-DPO: Precise and Efficient Post-Training for Diffusion Models

Recent advancements in diffusion models (DMs) have been propelled by alignment methods that post-train models to better conform to human preferences. However, these approaches typically require computation-intensive training of a base model and a reward model, which not only incurs substantial computational overhead but may also compromise model accuracy and training efficiency. To address these limitations, we propose Inversion-DPO, a novel alignment framework that circumvents reward modeling by reformulating Direct Preference Optimization (DPO) with DDIM inversion for DMs. Our method conducts intractable posterior sampling in Diffusion-DPO with the deterministic inversion from winning and losing samples to noise and thus derive a new post-training paradigm. This paradigm eliminates the need for auxiliary reward models or inaccurate appromixation, significantly enhancing both precision and efficiency of training. We apply Inversion-DPO to a basic task of text-to-image generation and a challenging task of compositional image generation. Extensive experiments show substantial performance improvements achieved by Inversion-DPO compared to existing post-training methods and highlight the ability of the trained generative models to generate high-fidelity compositionally coherent images. For the post-training of compostitional image geneation, we curate a paired dataset consisting of 11,140 images with complex structural annotations and comprehensive scores, designed to enhance the compositional capabilities of generative models. Inversion-DPO explores a new avenue for efficient, high-precision alignment in diffusion models, advancing their applicability to complex realistic generation tasks. Our code is available at https://github.com/MIGHTYEZ/Inversion-DPO

cs.CV↗

arXiv2025

DesignCoder: Hierarchy-Aware and Self-Correcting UI Code Generation with Large Language Models

Multimodal large language models (MLLMs) have streamlined front-end interface development by automating code generation. However, these models also introduce challenges in ensuring code quality. Existing approaches struggle to maintain both visual consistency and functional completeness in the generated components. Moreover, they lack mechanisms to assess the fidelity and correctness of the rendered pages. To address these issues, we propose DesignCoder, a novel hierarchical-aware and self-correcting automated code generation framework. Specifically, we introduce UI Grouping Chains, which enhance MLLMs' capability to understand and predict complex nested UI hierarchies. Subsequently, DesignCoder employs a hierarchical divide-and-conquer approach to generate front-end code. Finally, we incorporate a self-correction mechanism to improve the model's ability to identify and rectify errors in the generated code. Extensive evaluations on a dataset of UI mockups collected from both open-source communities and industry projects demonstrate that DesignCoder outperforms state-of-the-art baselines in React Native, a widely adopted UI framework. Our method achieves a 37.63%, 9.52%, 12.82% performance increase in visual similarity metrics (MSE, CLIP, SSIM) and significantly improves code structure similarity in terms of TreeBLEU, Container Match, and Tree Edit Distance by 30.19%, 29.31%, 24.67%. Furthermore, we conducted a user study with professional developers to assess the quality and practicality of the generated code. Results indicate that DesignCoder aligns with industry best practices, demonstrating high usability, readability, and maintainability. Our approach provides an efficient and practical solution for agile front-end development, enabling development teams to focus more on core functionality and product innovation.

cs.SE↗

arXiv2025

Diffusion Distillation With Direct Preference Optimization For Efficient 3D LiDAR Scene Completion

The application of diffusion models in 3D LiDAR scene completion is limited due to diffusion's slow sampling speed. Score distillation accelerates diffusion sampling but with performance degradation, while post-training with direct policy optimization (DPO) boosts performance using preference data. This paper proposes Distillation-DPO, a novel diffusion distillation framework for LiDAR scene completion with preference aligment. First, the student model generates paired completion scenes with different initial noises. Second, using LiDAR scene evaluation metrics as preference, we construct winning and losing sample pairs. Such construction is reasonable, since most LiDAR scene metrics are informative but non-differentiable to be optimized directly. Third, Distillation-DPO optimizes the student model by exploiting the difference in score functions between the teacher and student models on the paired completion scenes. Such procedure is repeated until convergence. Extensive experiments demonstrate that, compared to state-of-the-art LiDAR scene completion diffusion models, Distillation-DPO achieves higher-quality scene completion while accelerating the completion speed by more than 5-fold. Our method is the first to explore adopting preference learning in distillation to the best of our knowledge and provide insights into preference-aligned distillation. Our code is public available on https://github.com/happyw1nd/DistillationDPO.

cs.CV↗

arXiv2025

An Exploratory Study on How AI Awareness Impacts Human-AI Design Collaboration

The collaborative design process is intrinsically complicated and dynamic, and researchers have long been exploring how to enhance efficiency in this process. As Artificial Intelligence technology evolves, it has been widely used as a design tool and exhibited the potential as a design collaborator. Nevertheless, problems concerning how designers should communicate with AI in collaborative design remain unsolved. To address this research gap, we referred to how designers communicate fluently in human-human design collaboration, and found awareness to be an important ability for facilitating communication by understanding their collaborators and current situation. However, previous research mainly studied and supported human awareness, the possible impact AI awareness would bring to the human-AI collaborative design process, and the way to realize AI awareness remain unknown. In this study, we explored how AI awareness will impact human-AI collaboration through a Wizard-of-Oz experiment. Both quantitative and qualitative results supported that enabling AI to have awareness can enhance the communication fluidity between human and AI, thus enhancing collaboration efficiency. We further discussed the results and concluded design implications for future human-AI collaborative design systems.

cs.HC↗

arXiv2025

Integrating Sequence and Image Modeling in Irregular Medical Time Series Through Self-Supervised Learning

Medical time series are often irregular and face significant missingness, posing challenges for data analysis and clinical decision-making. Existing methods typically adopt a single modeling perspective, either treating series data as sequences or transforming them into image representations for further classification. In this paper, we propose a joint learning framework that incorporates both sequence and image representations. We also design three self-supervised learning strategies to facilitate the fusion of sequence and image representations, capturing a more generalizable joint representation. The results indicate that our approach outperforms seven other state-of-the-art models in three representative real-world clinical datasets. We further validate our approach by simulating two major types of real-world missingness through leave-sensors-out and leave-samples-out techniques. The results demonstrate that our approach is more robust and significantly surpasses other baselines in terms of classification performance.

cs.CV↗