arXiv subjects

Hongyu Wang

Publications and source records attributed to Hongyu Wang.

At least 55 records · Page 3Linked to original sources

arXiv2026

The Interdecadal Bipolar Oscillation: An Atmospheric Water Vapor Mode Driving Asynchronous Polar Climate Change

Climate change is progressing asynchronously between the Arctic and Antarctic, with important implications for global climate dynamics. While the Arctic has experienced rapid warming and pronounced amplification, the Antarctic has exhibited a delayed and heterogeneous response. Here, we identify an Interdecadal Bipolar Oscillation (IBO) in atmospheric water vapor, a coherent internal mode of variability that connects the two polar regions and helps explain their divergent climate trajectories over the past eight decades. Using reanalysis data alongside historical and pre-industrial control simulations from CMIP6, we demonstrate that the IBO is a robust internal variability mode with a quasi-period of 60 ~ 80 years. This oscillation modulates the background warming signal, with a phase shift in the late 1980s amplifying moistening and warming in the Arctic while concurrently suppressing changes in the Antarctic. Crucially, model projections suggest a possible phase reversal, which could slow water vapor increases in the Arctic while accelerating them in the Antarctic, potentially marking the start of rapid Antarctic climate change. While uncertainties remain in how climate models capture polar water vapor, our findings highlight the IBO as a pivotal driver of polar climate evolution and a potential contributor to emerging climate risks in the Southern Hemisphere.

physics.ao-ph↗

arXiv2026

MG-SLAM: Structure Gaussian Splatting SLAM with Manhattan World Hypothesis

Gaussian Splatting SLAMs have made significant advancements in improving the efficiency and fidelity of real-time reconstructions. However, these systems often encounter incomplete reconstructions in complex indoor environments, characterized by substantial holes due to unobserved geometry caused by obstacles or limited view angles. To address this challenge, we present Manhattan Gaussian SLAM, an RGB-D system that leverages the Manhattan World hypothesis to enhance geometric accuracy and completeness. By seamlessly integrating fused line segments derived from structured scenes, our method ensures robust tracking in textureless indoor areas. Moreover, The extracted lines and planar surface assumption allow strategic interpolation of new Gaussians in regions of missing geometry, enabling efficient scene completion. Extensive experiments conducted on both synthetic and real-world scenes demonstrate that these advancements enable our method to achieve state-of-the-art performance, marking a substantial improvement in the capabilities of Gaussian SLAM systems.

cs.RO↗

arXiv2026

MoTE: Mixture of Ternary Experts for Memory-efficient Large Multimodal Models

Large multimodal Mixture-of-Experts (MoEs) effectively scale the model size to boost performance while maintaining fixed active parameters. However, previous works primarily utilized full-precision experts during sparse up-cycling. Despite they show superior performance on end tasks, the large amount of experts introduces higher memory footprint, which poses significant challenges for the deployment on edge devices. In this work, we propose MoTE, a scalable and memory-efficient approach to train Mixture-of-Ternary-Experts models from dense checkpoint. Instead of training fewer high-precision experts, we propose to train more low-precision experts during up-cycling. Specifically, we use the pre-trained FFN as a shared expert and train ternary routed experts with parameters in {-1, 0, 1}. Extensive experiments show that our approach has promising scaling trend along model size. MoTE achieves comparable performance to full-precision baseline MoE-LLaVA while offering lower memory footprint. Furthermore, our approach is compatible with post-training quantization methods and the advantage further amplifies when memory-constraint goes lower. Given the same amount of expert memory footprint of 3.4GB and combined with post-training quantization, MoTE outperforms MoE-LLaVA by a gain of 4.3% average accuracy on end tasks, demonstrating its effectiveness and potential for memory-constrained devices.

cs.CV↗

arXiv2026

CMDAR: A Chinese Multi-scene Dynamic Audio Reasoning Benchmark with Diverse Challenges

The ability to reason from audio, including speech, environmental sounds, and music, is essential for AI agents to interact effectively in real-world scenarios. Existing benchmarks mainly focus on static or single-scene settings and English audio data and do not fully capture scenarios where multiple speakers, unfolding events, and heterogeneous audio sources interact. To address these challenges, we introduce CMDAR, a Chinese benchmark for evaluating models on complex, multi-scene, and dynamically evolving audio reasoning tasks. CMDAR comprises 3,000 carefully curated question-answer pairs linked to diverse audio clips, covering five categories of complex reasoning and spanning three question types. We benchmark 26 state-of-the-art audio language models on CMDAR and observe that they exhibit limitations in complex reasoning tasks. In CMDAR-main, Qwen2.5-Omni achieves 76.67% accuracy, whereas GPT-4o Audio reaches 68.47%. However, GPT-4o Audio substantially outperforms Qwen2.5-Omni on the more challenging multiple-choice with multiple audios and open-ended tasks. And we provide detail analysis corresponding suggestions for the future development of large audio language models.

cs.SD↗

arXiv2026

DenseSplat: Densifying Gaussian Splatting SLAM with Neural Radiance Prior

Gaussian SLAM systems excel in real-time rendering and fine-grained reconstruction compared to NeRF-based systems. However, their reliance on extensive keyframes is impractical for deployment in real-world robotic systems, which typically operate under sparse-view conditions that can result in substantial holes in the map. To address these challenges, we introduce DenseSplat, the first SLAM system that effectively combines the advantages of NeRF and 3DGS. DenseSplat utilizes sparse keyframes and NeRF priors for initializing primitives that densely populate maps and seamlessly fill gaps. It also implements geometry-aware primitive sampling and pruning strategies to manage granularity and enhance rendering efficiency. Moreover, DenseSplat integrates loop closure and bundle adjustment, significantly enhancing frame-to-frame tracking accuracy. Extensive experiments on multiple large-scale datasets demonstrate that DenseSplat achieves superior performance in tracking and mapping compared to current state-of-the-art methods.

cs.CV↗

arXiv2025

USE: A Unified Model for Universal Sound Separation and Extraction

Sound separation (SS) and target sound extraction (TSE) are fundamental techniques for addressing complex acoustic scenarios. While existing SS methods struggle with determining the unknown number of sound sources, TSE approaches require precisely specified clues to achieve optimal performance. This paper proposes a unified framework that synergistically combines SS and TSE to overcome their individual limitations. Our architecture employs two complementary components: 1) An Encoder-Decoder Attractor (EDA) network that automatically infers both the source count and corresponding acoustic clues for SS, and 2) A multi-modal fusion network that precisely interprets diverse user-provided clues (acoustic, semantic, or visual) for TSE. Through joint training with cross-task consistency constraints, we establish a unified latent space that bridges both paradigms. During inference, the system adaptively operates in either fully autonomous SS mode or clue-driven TSE mode. Experiments demonstrate remarkable performance in both tasks, with notable improvements of 1.4 dB SDR improvement in SS compared to baseline and 86\% TSE accuracy.

eess.AS↗

arXiv2025

RWKV-PCSSC: Exploring RWKV Model for Point Cloud Semantic Scene Completion

Semantic Scene Completion (SSC) aims to generate a complete semantic scene from an incomplete input. Existing approaches often employ dense network architectures with a high parameter count, leading to increased model complexity and resource demands. To address these limitations, we propose RWKV-PCSSC, a lightweight point cloud semantic scene completion network inspired by the Receptance Weighted Key Value (RWKV) mechanism. Specifically, we introduce a RWKV Seed Generator (RWKV-SG) module that can aggregate features from a partial point cloud to produce a coarse point cloud with coarse features. Subsequently, the point-wise feature of the point cloud is progressively restored through multiple stages of the RWKV Point Deconvolution (RWKV-PD) modules. By leveraging a compact and efficient design, our method achieves a lightweight model representation. Experimental results demonstrate that RWKV-PCSSC reduces the parameter count by 4.18$\times$ and improves memory efficiency by 1.37$\times$ compared to state-of-the-art methods PointSSC. Furthermore, our network achieves state-of-the-art performance on established indoor (SSC-PC, NYUCAD-PC) and outdoor (PointSSC) scene dataset, as well as on our proposed datasets (NYUCAD-PC-V2, 3D-FRONT-PC).

cs.CV↗

arXiv2025

STG-Avatar: Animatable Human Avatars via Spacetime Gaussian

Realistic animatable human avatars from monocular videos are crucial for advancing human-robot interaction and enhancing immersive virtual experiences. While recent research on 3DGS-based human avatars has made progress, it still struggles with accurately representing detailed features of non-rigid objects (e.g., clothing deformations) and dynamic regions (e.g., rapidly moving limbs). To address these challenges, we present STG-Avatar, a 3DGS-based framework for high-fidelity animatable human avatar reconstruction. Specifically, our framework introduces a rigid-nonrigid coupled deformation framework that synergistically integrates Spacetime Gaussians (STG) with linear blend skinning (LBS). In this hybrid design, LBS enables real-time skeletal control by driving global pose transformations, while STG complements it through spacetime adaptive optimization of 3D Gaussians. Furthermore, we employ optical flow to identify high-dynamic regions and guide the adaptive densification of 3D Gaussians in these regions. Experimental results demonstrate that our method consistently outperforms state-of-the-art baselines in both reconstruction quality and operational efficiency, achieving superior quantitative metrics while retaining real-time rendering capabilities. Our code is available at https://github.com/jiangguangan/STG-Avatar

cs.CV↗

arXiv2025

Pure Core Sets of $n \times n$ Matrices over Finite Fields

This paper studies the structure of core sets under different similarity classes. We investigate the influence of factors of the minimal polynomial with different degrees on the structure of core sets. When $F$ is a finite field of prime order, we study the upper bound on the size of a non-core set in a similarity class in $M_n(F)$. We prove that as $|F|$ increases, the proportion of pure core sets among subsets of $M_n(F)$ tends to $1$.

math.RA↗

arXiv2025

Online Specific Emitter Identification via Collision-Alleviated Signal Hash

Specific Emitter Identification (SEI) has been widely studied, aiming to distinguish signals from different emitters given training samples from those emitters. However, real-world scenarios often require identifying signals from novel emitters previously unseen. Since these novel emitters only have a few or no prior samples, existing models struggle to identify signals from novel emitters online and tend to bias toward the distribution of seen emitters. To address these challenges, we propose the Online Specific Emitter Identification (OSEI) task, comprising both online \revise{few-shot and generalized zero-shot} learning tasks. It requires constructing models using signal samples from seen emitters and then identifying new samples from seen and novel emitters online during inference. We propose a novel hash-based model, Collision-Alleviated Signal Hash (CASH), providing a unified approach for addressing the OSEI task. The CASH operates in two steps: in the seen emitters identifying step, a signal encoder and a seen emitters identifier determine whether the signal sample is from seen emitters, mitigating the model from biasing toward seen emitters distribution. In the signal hash coding step, an online signal hasher assigns a hash code to each signal sample, identifying its specific emitter. Experimental results on real-world signal datasets (i.e., ADSB and ORACLE) demonstrate that our method accurately identifies signals from both seen and novel emitters online. This model outperforms existing methods by a minimum of 6.08\% and 8.55\% in accuracy for the few-shot and \revise{generalized zero-shot learning }tasks, respectively. The code will be open-sourced at \href{https://github.com/IntelliSensing/OSEI-CASH}{https://github.com/IntelliSensing/OSEI-CASH}.

eess.SP↗

arXiv2025

MeanSE: Efficient Generative Speech Enhancement with Mean Flows

Speech enhancement (SE) improves degraded speech's quality, with generative models like flow matching gaining attention for their outstanding perceptual quality. However, the flow-based model requires multiple numbers of function evaluations (NFEs) to achieve stable and satisfactory performance, leading to high computational load and poor 1-NFE performance. In this paper, we propose MeanSE, an efficient generative speech enhancement model using mean flows, which models the average velocity field to achieve high-quality 1-NFE enhancement. Experimental results demonstrate that our proposed MeanSE significantly outperforms the flow matching baseline with a single NFE, exhibiting extremely better out-of-domain generalization capabilities.

eess.AS↗

arXiv2025

BitNet v2: Native 4-bit Activations with Hadamard Transformation for 1-bit LLMs

Efficient deployment of 1-bit Large Language Models (LLMs) is hindered by activation outliers, which complicate quantization to low bit-widths. We introduce BitNet v2, a novel framework enabling native 4-bit activation quantization for 1-bit LLMs. To tackle outliers in attention and feed-forward network activations, we propose H-BitLinear, a module applying an online Hadamard transformation prior to activation quantization. This transformation smooths sharp activation distributions into more Gaussian-like forms, suitable for low-bit representation. Experiments show BitNet v2 trained from scratch with 8-bit activations matches BitNet b1.58 performance. Crucially, BitNet v2 achieves minimal performance degradation when trained with native 4-bit activations, significantly reducing memory footprint and computational cost for batched inference.

cs.CL↗

arXiv2025

Dy3DGS-SLAM: Monocular 3D Gaussian Splatting SLAM for Dynamic Environments

Current Simultaneous Localization and Mapping (SLAM) methods based on Neural Radiance Fields (NeRF) or 3D Gaussian Splatting excel in reconstructing static 3D scenes but struggle with tracking and reconstruction in dynamic environments, such as real-world scenes with moving elements. Existing NeRF-based SLAM approaches addressing dynamic challenges typically rely on RGB-D inputs, with few methods accommodating pure RGB input. To overcome these limitations, we propose Dy3DGS-SLAM, the first 3D Gaussian Splatting (3DGS) SLAM method for dynamic scenes using monocular RGB input. To address dynamic interference, we fuse optical flow masks and depth masks through a probabilistic model to obtain a fused dynamic mask. With only a single network iteration, this can constrain tracking scales and refine rendered geometry. Based on the fused dynamic mask, we designed a novel motion loss to constrain the pose estimation network for tracking. In mapping, we use the rendering loss of dynamic pixels, color, and depth to eliminate transient interference and occlusion caused by dynamic objects. Experimental results demonstrate that Dy3DGS-SLAM achieves state-of-the-art tracking and rendering in dynamic environments, outperforming or matching existing RGB-D methods.

cs.CV↗

arXiv2025

Learning dissection trajectories from expert surgical videos via imitation learning with equivariant diffusion

Endoscopic Submucosal Dissection (ESD) is a well-established technique for removing epithelial lesions. Predicting dissection trajectories in ESD videos offers significant potential for enhancing surgical skill training and simplifying the learning process, yet this area remains underexplored. While imitation learning has shown promise in acquiring skills from expert demonstrations, challenges persist in handling uncertain future movements, learning geometric symmetries, and generalizing to diverse surgical scenarios. To address these, we introduce a novel approach: Implicit Diffusion Policy with Equivariant Representations for Imitation Learning (iDPOE). Our method models expert behavior through a joint state action distribution, capturing the stochastic nature of dissection trajectories and enabling robust visual representation learning across various endoscopic views. By incorporating a diffusion model into policy learning, iDPOE ensures efficient training and sampling, leading to more accurate predictions and better generalization. Additionally, we enhance the model's ability to generalize to geometric symmetries by embedding equivariance into the learning process. To address state mismatches, we develop a forward-process guided action inference strategy for conditional sampling. Using an ESD video dataset of nearly 2000 clips, experimental results show that our approach surpasses state-of-the-art methods, both explicit and implicit, in trajectory prediction. To the best of our knowledge, this is the first application of imitation learning to surgical skill development for dissection trajectory prediction.

cs.CV↗

arXiv2025

M4U: Evaluating Multilingual Understanding and Reasoning for Large Multimodal Models

Multilingual capability is an essential aspect for large multimodal models, since they are usually deployed across various countries and languages. However, most existing benchmarks for multilingual multimodal reasoning struggle to differentiate between models of varying performance; even language models without visual capabilities can easily achieve high scores. This leaves a comprehensive evaluation of leading multilingual multimodal models largely unexplored. In this work, we introduce M4U, a novel and challenging benchmark for assessing the capability of multi-discipline multilingual multimodal understanding and reasoning. M4U contains 10k samples covering 64 disciplines across 16 subfields in Science, Engineering, and Healthcare in six languages. Using M4U, we conduct extensive evaluations of leading Large Multimodal Models (LMMs) and Large Language Models (LLMs) with external tools. The evaluation results demonstrate that the state-of-the-art model, GPT-4o, achieves only 47.6% average accuracy on M4U. Additionally, we observe that the leading LMMs exhibit significant language preferences. Our in-depth analysis indicates that leading LMMs, including GPT-4o, struggle to perform reasoning using multilingual information present in both visual and textual context. Specifically, they suffer performance degradation when prompted with cross-lingual multimodal questions. Our code and dataset is public available.

cs.CV↗

arXiv2025

BitNet b1.58 2B4T Technical Report

We introduce BitNet b1.58 2B4T, the first open-source, native 1-bit Large Language Model (LLM) at the 2-billion parameter scale. Trained on a corpus of 4 trillion tokens, the model has been rigorously evaluated across benchmarks covering language understanding, mathematical reasoning, coding proficiency, and conversational ability. Our results demonstrate that BitNet b1.58 2B4T achieves performance on par with leading open-weight, full-precision LLMs of similar size, while offering significant advantages in computational efficiency, including substantially reduced memory footprint, energy consumption, and decoding latency. To facilitate further research and adoption, the model weights are released via Hugging Face along with open-source inference implementations for both GPU and CPU architectures.

cs.CL↗

arXiv2025

NGM-SLAM: Gaussian Splatting SLAM with Radiance Field Submap

SLAM systems based on Gaussian Splatting have garnered attention due to their capabilities for rapid real-time rendering and high-fidelity mapping. However, current Gaussian Splatting SLAM systems usually struggle with large scene representation and lack effective loop closure detection. To address these issues, we introduce NGM-SLAM, the first 3DGS based SLAM system that utilizes neural radiance field submaps for progressive scene expression, effectively integrating the strengths of neural radiance fields and 3D Gaussian Splatting. We utilize neural radiance field submaps as supervision and achieve high-quality scene expression and online loop closure adjustments through Gaussian rendering of fused submaps. Our results on multiple real-world scenes and large-scale scene datasets demonstrate that our method can achieve accurate hole filling and high-quality scene expression, supporting monocular, stereo, and RGB-D inputs, and achieving state-of-the-art scene reconstruction and tracking performance.

cs.RO↗

arXiv2025

A Revisit of Large-Scale Patterns in Middle Stratospheric Circulation Variations

Variations in stratospheric atmospheric circulation significantly influence tropospheric weather and climate, and understanding these variations can guide stratospheric aircraft development and operations. Despite a century of progress, large-scale patterns in stratospheric circulation remain poorly understood due to the stratosphere's complex nature. To address this, we applied the eigen microstate approach (EMA) to analyze zonal wind from 70-10 hPa using ERA5 reanalysis data from 1980-2022. We focused on the three leading modes, corresponding to the quasi-biennial oscillation (QBO) and stratospheric circulation in the Arctic and Antarctic. After removing high-frequency components, we observed a significant 11-year cycle in the Antarctic stratospheric circulation mode, possibly linked to the solar cycle. In contrast, the Arctic mode showed a 5-6-year cycle without 11-year periodicity. This difference likely arises from the seasonal timing of polar vortex breakdowns: the Antarctic vortex persists into late spring and summer, making it more sensitive to solar radiation, while the Arctic vortex peaks in winter and early spring. The fourth mode showed features of a Southern Hemisphere dipole and was significantly correlated with the Antarctic mode, leading it by about two months. Finally, we developed a linear prediction model that demonstrated predictive skill for the Antarctic polar vortex.

physics.ao-ph↗