arXiv subjects

Ruoyu Wang

Publications and source records attributed to Ruoyu Wang.

At least 73 records · Page 4Linked to original sources

arXiv2025

From Semantics, Scene to Instance-awareness: Distilling Foundation Model for Grounded Open-vocabulary Situation Recognition

Recent Multimodal Large Language Models (MLLMs) exhibit strong zero-shot abilities but struggle with complex Grounded Situation Recognition (GSR) and are resource-intensive for edge device deployment. Meanwhile, conventional GSR models often lack generalization ability, falling short in recognizing unseen and rare situations. In this paper, we exploit transferring knowledge from a teacher MLLM to a small GSR model to enhance its generalization and zero-shot abilities, thereby introducing the task of Open-vocabulary Grounded Situation Recognition (Ov-GSR). To achieve this, we propose Multimodal Interactive Prompt Distillation (MIPD), a novel framework that distills enriched multimodal knowledge from the foundation model, enabling the student Ov-GSR model to recognize unseen situations and be better aware of rare situations. Specifically, the MIPD framework first leverages the LLM-based Judgmental Rationales Generator (JRG) to construct positive and negative glimpse and gaze rationales enriched with contextual semantic information. The proposed scene-aware and instance-perception prompts are then introduced to align rationales with visual information from the MLLM teacher via the Negative-Guided Multimodal Prompting Alignment (NMPA) module, effectively capturing holistic and perceptual multimodal knowledge. Finally, the aligned multimodal knowledge is distilled into the student Ov-GSR model, providing a stronger foundation for generalization that enhances situation understanding, bridges the gap between seen and unseen scenarios, and mitigates prediction bias in rare cases. We evaluate MIPD on the refined Ov-SWiG dataset, achieving superior performance on seen, rare, and unseen situations, and further demonstrate improved unseen detection on the HICO-DET dataset.

cs.CV↗

arXiv2025

A decomposition method in the multivariate feedback particle filter via tensor product Hermite polynomials

The feedback particle filter (FPF), a resampling-free algorithm proposed over a decade ago, modifies the particle filter (PF) by incorporating a feedback structure. Each particle in FPF is regulated via a feedback gain function (lacking a closed-form expression), which solves a Poisson's equation with a probability-weighted Laplacian. While approximate solutions to this equation have been extensively studied in recent literature, no efficient multivariate algorithm exists. In this paper, we focus on the decomposition method for multivariate gain functions in FPF, which has been proven efficient for scalar FPF with polynomial observation functions. Its core is splitting the Poisson's equation into two exactly solvable sub-equations. Key challenges in extending it to multivariate FPF include ensuring the invertibility of the coefficient matrix in one sub-equation and constructing a weighted-radial solution in the other. The proposed method's computational complexity grows at most polynomially with the state dimension, a dramatic improvement over the exponential growth of most particle-based algorithms. Numerical experiments compare the decomposition method with traditional methods: the extended Kalman filter (EKF), PF, and FPF with constant-gain or kernel-based gain approximations. Results show it outperforms PF and FPF with other gain approximations in both accuracy and efficiency, achieving the shortest CPU time among methods with comparable performance.

math.OC↗

arXiv2025

Adaptive Stochastic Coefficients for Accelerating Diffusion Sampling

Diffusion-based generative processes, formulated as differential equation solving, frequently balance computational speed with sample quality. Our theoretical investigation of ODE- and SDE-based solvers reveals complementary weaknesses: ODE solvers accumulate irreducible gradient error along deterministic trajectories, while SDE methods suffer from amplified discretization errors when the step budget is limited. Building upon this insight, we introduce AdaSDE, a novel single-step SDE solver that aims to unify the efficiency of ODEs with the error resilience of SDEs. Specifically, we introduce a single per-step learnable coefficient, estimated via lightweight distillation, which dynamically regulates the error correction strength to accelerate diffusion sampling. Notably, our framework can be integrated with existing solvers to enhance their capabilities. Extensive experiments demonstrate state-of-the-art performance: at 5 NFE, AdaSDE achieves FID scores of 4.18 on CIFAR-10, 8.05 on FFHQ and 6.96 on LSUN Bedroom. Codes are available in https://github.com/WLU-wry02/AdaSDE.

cs.CV↗

arXiv2025

Spatially Aware Self-Supervised Models for Multi-Channel Neural Speaker Diarization

Self-supervised models such as WavLM have demonstrated strong performance for neural speaker diarization. However, these models are typically pre-trained on single-channel recordings, limiting their effectiveness in multi-channel scenarios. Existing diarization systems built on these models often rely on DOVER-Lap to combine outputs from individual channels. Although effective, this approach incurs substantial computational overhead and fails to fully exploit spatial information. In this work, building on DiariZen, a pipeline that combines WavLM-based local endto-end neural diarization with speaker embedding clustering, we introduce a lightweight approach to make pre-trained WavLM spatially aware by inserting channel communication modules into the early layers. Our method is agnostic to both the number of microphone channels and array topologies, ensuring broad applicability. We further propose to fuse multi-channel speaker embeddings by leveraging spatial attention weights. Evaluations on five public datasets show consistent improvements over single-channel baselines and demonstrate superior performance and efficiency compared with DOVER-Lap. Our source code is publicly available at https://github.com/BUTSpeechFIT/DiariZen.

eess.AS↗

arXiv2025

Identification and Estimation of Heterogeneous Interference Effects under Unknown Network

Interference--in which a unit's outcome is affected by the treatment of other units--poses significant challenges for the identification and estimation of causal effects. Most existing methods for estimating interference effects assume that the interference networks are known. In many practical settings, this assumption is unrealistic as such networks are typically latent. To address this challenge, we propose a novel framework for identifying and estimating heterogeneous group-level interference effects without requiring a known interference network. Specifically, we assume a shared latent community structure between the observed network and the unknown interference network. We demonstrate that interference effects are identifiable if and only if group-level interference effects are heterogeneous, and we establish the consistency and asymptotic normality of the maximum likelihood estimator (MLE). To handle the intractable likelihood function and facilitate the computation, we propose a Bayesian implementation and show that the posterior concentrates around the MLE. A series of simulation studies demonstrate the effectiveness of the proposed method and its superior performance compared with competitors. We apply our proposed framework to the encounter data of stroke patients from the California Department of Healthcare Access and Information (HCAI) and evaluate the causal interference effects of certain intervention in one hospital on the outcomes of other hospitals.

stat.ME↗

arXiv2025

Subsampled One-Step Estimation for Fast Statistical Inference

Subsampling is an effective approach to alleviate the computational burden associated with large-scale datasets. Nevertheless, existing subsampling estimators incur a substantial loss in estimation efficiency compared to estimators based on the full dataset. Specifically, the convergence rate of existing subsampling estimators is typically $n^{-1/2}$ rather than $N^{-1/2}$, where $n$ and $N$ denote the subsample and full data sizes, respectively. This paper proposes a subsampled one-step (SOS) method to mitigate the estimation efficiency loss utilizing the asymptotic expansions of the subsampling and full-data estimators. The resulting SOS estimator is computationally efficient and achieves a fast convergence rate of $\max\{n^{-1}, N^{-1/2}\}$ rather than $n^{-1/2}$. We establish the asymptotic distribution of the SOS estimator, which can be non-normal in general, and construct confidence intervals on top of the asymptotic distribution. Furthermore, we prove that the SOS estimator is asymptotically normal and equivalent to the full data-based estimator when $n / \sqrt{N} \to \infty$.Simulation studies and real data analyses were conducted to demonstrate the finite sample performance of the SOS estimator. Numerical results suggest that the SOS estimator is almost as computationally efficient as the uniform subsampling estimator while achieving similar estimation efficiency to the full data-based estimator.

stat.ME↗

arXiv2025

Causal Effect Identification and Inference with Endogenous Exposures and a Light-tailed Error

Endogeneity poses significant challenges in causal inference across various research domains. This paper proposes a novel approach to identify and estimate causal effects in the presence of endogeneity. We consider a structural equation with endogenous exposures and an additive error term. Assuming the light-tailedness of the error term, we show that the causal effect can be identified by contrasting extreme conditional quantiles of the outcome given the exposures. Unlike many existing results, our identification approach does not rely on additional parametric assumptions or auxiliary variables. Building on the identification result, we develop a new method that estimates the causal effect using extreme quantile regression. We establish the consistency of the proposed extreme-based estimator under a general additive structural equation and demonstrate its asymptotic normality in the linear model setting. Simulations and data analysis of an automobile sale dataset show the effectiveness of our method in handling endogeneity.

stat.ME↗

arXiv2025

SimGenHOI: Physically Realistic Whole-Body Humanoid-Object Interaction via Generative Modeling and Reinforcement Learning

Generating physically realistic humanoid-object interactions (HOI) is a fundamental challenge in robotics. Existing HOI generation approaches, such as diffusion-based models, often suffer from artifacts such as implausible contacts, penetrations, and unrealistic whole-body actions, which hinder successful execution in physical environments. To address these challenges, we introduce SimGenHOI, a unified framework that combines the strengths of generative modeling and reinforcement learning to produce controllable and physically plausible HOI. Our HOI generative model, based on Diffusion Transformers (DiT), predicts a set of key actions conditioned on text prompts, object geometry, sparse object waypoints, and the initial humanoid pose. These key actions capture essential interaction dynamics and are interpolated into smooth motion trajectories, naturally supporting long-horizon generation. To ensure physical realism, we design a contact-aware whole-body control policy trained with reinforcement learning, which tracks the generated motions while correcting artifacts such as penetration and foot sliding. Furthermore, we introduce a mutual fine-tuning strategy, where the generative model and the control policy iteratively refine each other, improving both motion realism and tracking robustness. Extensive experiments demonstrate that SimGenHOI generates realistic, diverse, and physically plausible humanoid-object interactions, achieving significantly higher tracking success rates in simulation and enabling long-horizon manipulation tasks. Code will be released upon acceptance on our project page: https://xingxingzuo.github.io/simgen_hoi.

cs.RO↗

arXiv2025

The Silent Assistant: NoiseQuery as Implicit Guidance for Goal-Driven Image Generation

In this work, we introduce NoiseQuery as a novel method for enhanced noise initialization in versatile goal-driven text-to-image (T2I) generation. Specifically, we propose to leverage an aligned Gaussian noise as implicit guidance to complement explicit user-defined inputs, such as text prompts, for better generation quality and controllability. Unlike existing noise optimization methods designed for specific models, our approach is grounded in a fundamental examination of the generic finite-step noise scheduler design in diffusion formulation, allowing better generalization across different diffusion-based architectures in a tuning-free manner. This model-agnostic nature allows us to construct a reusable noise library compatible with multiple T2I models and enhancement techniques, serving as a foundational layer for more effective generation. Extensive experiments demonstrate that NoiseQuery enables fine-grained control and yields significant performance boosts not only over high-level semantics but also over low-level visual attributes, which are typically difficult to specify through text alone, with seamless integration into current workflows with minimal computational overhead.

cs.CV↗

arXiv2025

DICE: Dynamic In-Context Example Selection in LLM Agents via Efficient Knowledge Transfer

Large language model-based agents, empowered by in-context learning (ICL), have demonstrated strong capabilities in complex reasoning and tool-use tasks. However, existing works have shown that the effectiveness of ICL is highly sensitive to the choice of demonstrations, with suboptimal examples often leading to unstable or degraded performance. While prior work has explored example selection, including in some agentic or multi-step settings, existing approaches typically rely on heuristics or task-specific designs and lack a general, theoretically grounded criterion for what constitutes an effective demonstration across reasoning steps. Therefore, it is non-trivial to develop a principled, general-purpose method for selecting demonstrations that consistently benefit agent performance. In this paper, we address this challenge with DICE, Dynamic In-Context Example Selection for LLM Agents, a theoretically grounded ICL framework for agentic tasks that selects the most relevant demonstrations at each step of reasoning. Our approach decomposes demonstration knowledge into transferable and non-transferable components through a causal lens, showing how the latter can introduce spurious dependencies that impair generalization. We further propose a stepwise selection criterion with a formal guarantee of improved agent performance. Importantly, DICE is a general, framework-agnostic solution that can be integrated as a plug-in module into existing agentic frameworks without any additional training cost. Extensive experiments across diverse domains demonstrate our method's effectiveness and generality, highlighting the importance of principled, context-aware demo selection for robust and efficient LLM agents.

cs.AI↗

arXiv2025

Latent Swap Joint Diffusion for 2D Long-Form Latent Generation

This paper introduces Swap Forward (SaFa), a modality-agnostic and efficient method to generate seamless and coherence long spectrum and panorama through latent swap joint diffusion across multi-views. We first investigate the spectrum aliasing problem in spectrum-based audio generation caused by existing joint diffusion methods. Through a comparative analysis of the VAE latent representation of Mel-spectra and RGB images, we identify that the failure arises from excessive suppression of high-frequency components during the spectrum denoising process due to the averaging operator. To address this issue, we propose Self-Loop Latent Swap, a frame-level bidirectional swap applied to the overlapping region of adjacent views. Leveraging stepwise differentiated trajectories of adjacent subviews, this swap operator adaptively enhances high-frequency components and avoid spectrum distortion. Furthermore, to improve global cross-view consistency in non-overlapping regions, we introduce Reference-Guided Latent Swap, a unidirectional latent swap operator that provides a centralized reference trajectory to synchronize subview diffusions. By refining swap timing and intervals, we can achieve a cross-view similarity-diversity balance in a forward-only manner. Quantitative and qualitative experiments demonstrate that SaFa significantly outperforms existing joint diffusion methods and even training-based methods in audio generation using both U-Net and DiT models, along with effective longer length adaptation. It also adapts well to panorama generation, achieving comparable performance with 2 $\sim$ 20 $\times$ faster speed and greater model generalizability. More generation demos are available at https://swapforward.github.io/

cs.SD↗

arXiv2025

Causality-aligned Prompt Learning via Diffusion-based Counterfactual Generation

Prompt learning has garnered attention for its efficiency over traditional model training and fine-tuning. However, existing methods, constrained by inadequate theoretical foundations, encounter difficulties in achieving causally invariant prompts, ultimately falling short of capturing robust features that generalize effectively across categories. To address these challenges, we introduce the $\textit{\textbf{DiCap}}$ model, a theoretically grounded $\textbf{Di}$ffusion-based $\textbf{C}$ounterf$\textbf{a}$ctual $\textbf{p}$rompt learning framework, which leverages a diffusion process to iteratively sample gradients from the marginal and conditional distributions of the causal model, guiding the generation of counterfactuals that satisfy the minimal sufficiency criterion. Grounded in rigorous theoretical derivations, this approach guarantees the identifiability of counterfactual outcomes while imposing strict bounds on estimation errors. We further employ a contrastive learning framework that leverages the generated counterfactuals, thereby enabling the refined extraction of prompts that are precisely aligned with the causal features of the data. Extensive experimental results demonstrate that our method performs excellently across tasks such as image classification, image-text retrieval, and visual question answering, with particularly strong advantages in unseen categories.

cs.AI↗

arXiv2025

LongSplat: Online Generalizable 3D Gaussian Splatting from Long Sequence Images

3D Gaussian Splatting achieves high-fidelity novel view synthesis, but its application to online long-sequence scenarios is still limited. Existing methods either rely on slow per-scene optimization or fail to provide efficient incremental updates, hindering continuous performance. In this paper, we propose LongSplat, an online real-time 3D Gaussian reconstruction framework designed for long-sequence image input. The core idea is a streaming update mechanism that incrementally integrates current-view observations while selectively compressing redundant historical Gaussians. Crucial to this mechanism is our Gaussian-Image Representation (GIR), a representation that encodes 3D Gaussian parameters into a structured, image-like 2D format. GIR simultaneously enables efficient fusion of current-view and historical Gaussians and identity-aware redundancy compression. These functions enable online reconstruction and adapt the model to long sequences without overwhelming memory or computational costs. Furthermore, we leverage an existing image compression method to guide the generation of more compact and higher-quality 3D Gaussians. Extensive evaluations demonstrate that LongSplat achieves state-of-the-art efficiency-quality trade-offs in real-time novel view synthesis, delivering real-time reconstruction while reducing Gaussian counts by 44\% compared to existing per-pixel Gaussian prediction methods.

cs.CV↗

arXiv2025

Distilling Parallel Gradients for Fast ODE Solvers of Diffusion Models

Diffusion models (DMs) have achieved state-of-the-art generative performance but suffer from high sampling latency due to their sequential denoising nature. Existing solver-based acceleration methods often face image quality degradation under a low-latency budget. In this paper, we propose the Ensemble Parallel Direction solver (dubbed as \ours), a novel ODE solver that mitigates truncation errors by incorporating multiple parallel gradient evaluations in each ODE step. Importantly, since the additional gradient computations are independent, they can be fully parallelized, preserving low-latency sampling. Our method optimizes a small set of learnable parameters in a distillation fashion, ensuring minimal training overhead. In addition, our method can serve as a plugin to improve existing ODE samplers. Extensive experiments on various image synthesis benchmarks demonstrate the effectiveness of our \ours~in achieving high-quality and low-latency sampling. For example, at the same latency level of 5 NFE, EPD achieves an FID of 4.47 on CIFAR-10, 7.97 on FFHQ, 8.17 on ImageNet, and 8.26 on LSUN Bedroom, surpassing existing learning-based solvers by a significant margin. Codes are available in https://github.com/BeierZhu/EPD.

cs.CV↗

arXiv2025

Uncovering coupled ionic-polaronic dynamics and interfacial enhancement in Li$_x$FePO$_4$

Understanding and controlling coupled ionic-polaronic dynamics is crucial for optimizing electrochemical performance in battery materials. However, studying such coupled dynamics remains challenging due to the intricate interplay between Li-ion configurations, polaron charge ordering, and lattice vibrations. Here, we develop a fine-tuned machine-learned force field (MLFF) for Li$_x$FePO$_4$ that captures coupled ion-polaron behavior. Our simulations reveal picosecond-scale polaron flips occurring orders of magnitude faster than Li-ion migration, featuring strong correlation to Li configurations. Notably, polaron charge fluctuations are further enhanced at Li-rich/Li-poor phase boundaries, suggesting a potential interfacial electronic conduction mechanism. These results demonstrate the capability of fine-tuned MLFFs to resolve complex coupled transport and provide insight into emergent ionic-polaronic dynamics in multivalent battery cathodes.

cond-mat.mtrl-sci↗

arXiv2025

Broadband nonlinear optical microresonator array for topological second harmonic generation

Topological photonics enables robust light manipulation with third-order optical nonlinearity, yet integrating second-order optical nonlinearity into a topological system faces fundamental challenges: frequency-dependent topological bandgaps impede simultaneous edge states for pump and second harmonic photons at an octave. Here we present a broadband topological nonlinear photonic system via dual frequency topological bandgap engineering in a 2D nonlinear microresonator array. By designing a square lattice with synthetic magnetic fluxes, we achieve topological phase matching preserving unidirectional edge states for both frequencies while enabling efficient second-harmonic generation. The system exhibits flux-programmable SH chirality, where SH photons reverse propagation direction via Chern number transitions (e.g. C = -1 to +1) without sacrificing robustness. Moreover, we show that the design theoretically yields over 100 times higher SHG efficiency than single resonators at high powers via topology-enhanced coherent buildup. Our topological SHG works in a parameter regime that can be readily accessed by using existing low-loss integrated photon platforms like thin film lithium niobite.

physics.optics↗

arXiv2025

Piezoelectric-Metal Phononic Crystal Enabling GHz Tunable Ultrahigh $Q$ Quasi-BIC mode

The integration of GHz-frequency, high quality factor ($Q$), and electrically tunable acoustic resonators holds significant potential for advancing applications in quantum information technologies, microwave photonics, and reconfigurable RF systems. However, simultaneously achieving these three characteristics within a single, scalable platform remains a fundamental challenge. Here, we report the experimental demonstration of a GHz quasi-BIC resonator in a piezoelectric thin-film shear horizontal (SH) wave system, achieved through a structurally simple piezoelectric-metal phononic crystal (PnC) architecture on a LiNbO$_3$ thin film. This approach enables leaky Fabry-Perot coupling mode and localized trapping quasi-BIC mode. Without the need for deep etching or intricate patterning, we achieve a room-temperature quality factor of $6\times 10^4$ at ~1 GHz in ambient air, corresponding to an $f\times Q$ product of $6\times 10^{13}$ Hz at quasi-BIC mode. Furthermore, we demonstrate efficient electrical tunability via low-voltage (0.6 V) electrothermal modulation of the PnC structure, enabling a reversible transition between trapped and transmission states and yielding a high-contrast amplitude modulation of 47.75 dB. Our results establish a lithography-friendly, fabrication-tolerant platform for realizing tunable, high-$Q$ acoustic resonators at GHz frequencies, overcoming longstanding barriers in phononic device engineering. This work opens new directions for scalable on-chip phononic circuits in quantum acoustics, reconfigurable RF systems, and signal processing applications.

physics.app-ph↗

arXiv2025

Efficient and Generalizable Environmental Understanding for Visual Navigation

Visual Navigation is a core task in Embodied AI, enabling agents to navigate complex environments toward given objectives. Across diverse settings within Navigation tasks, many necessitate the modelling of sequential data accumulated from preceding time steps. While existing methods perform well, they typically process all historical observations simultaneously, overlooking the internal association structure within the data, which may limit the potential for further improvements in task performance. We address this by examining the unique characteristics of Navigation tasks through the lens of causality, introducing a causal framework to highlight the limitations of conventional sequential methods. Leveraging this insight, we propose Causality-Aware Navigation (CAN), which incorporates a Causal Understanding Module to enhance the agent's environmental understanding capability. Empirical evaluations show that our approach consistently outperforms baselines across various tasks and simulation environments. Extensive ablations studies attribute these gains to the Causal Understanding Module, which generalizes effectively in both Reinforcement and Supervised Learning settings without computational overhead.

cs.AI↗