Yuke Li

Hao Ni, Yuke Li, Lianli Gao et al.

Domain generalization person re-identification (DG-ReID) aims to train a model on source domains and generalize well on unseen domains. Vision Transformer usually yields better generalization ability than common CNN networks under distribution shifts. However, Transformer-based ReID models inevitably over-fit to domain-specific biases due to the supervised learning strategy on the source domain. We observe that while the global images of different IDs should have different features, their similar local parts (e.g., black backpack) are not bounded by this constraint. Motivated by this, we propose a pure Transformer model (termed Part-aware Transformer) for DG-ReID by designing a proxy task, named Cross-ID Similarity Learning (CSL), to mine local visual information shared by different IDs. This proxy task allows the model to learn generic features because it only cares about the visual similarity of the parts regardless of the ID labels, thus alleviating the side effect of domain-specific biases. Based on the local similarity obtained in CSL, a Part-guided Self-Distillation (PSD) is proposed to further improve the generalization of global features. Our method achieves state-of-the-art performance under most DG ReID settings. Under the Market$\to$Duke setting, our method exceeds state-of-the-art by 10.9% and 12.8% in Rank1 and mAP, respectively. The code is available at https://github.com/liyuke65535/Part-Aware-Transformer.

Yuhang Liu, Boyi Sun, Yuke Li et al.

To develop the next generation of intelligent LiDARs, we propose a novel framework of parallel LiDARs and construct a hardware prototype in our experimental platform, DAWN (Digital Artificial World for Natural). It emphasizes the tight integration of physical and digital space in LiDAR systems, with networking being one of its supported core features. In the context of autonomous driving, V2V (Vehicle-to-Vehicle) technology enables efficient information sharing between different agents which significantly promotes the development of LiDAR networks. However, current research operates under an ideal situation where all vehicles are equipped with identical LiDAR, ignoring the diversity of LiDAR categories and operating frequencies. In this paper, we first utilize OpenCDA and RLS (Realistic LiDAR Simulation) to construct a novel heterogeneous LiDAR dataset named OPV2V-HPL. Additionally, we present HPL-ViT, a pioneering architecture designed for robust feature fusion in heterogeneous and dynamic scenarios. It uses a graph-attention Transformer to extract domain-specific features for each agent, coupled with a cross-attention mechanism for the final fusion. Extensive experiments on OPV2V-HPL demonstrate that HPL-ViT achieves SOTA (state-of-the-art) performance in all settings and exhibits outstanding generalization capabilities.

Zhanwen Liu, Nan Yang, Yang Wang et al.

Traffic object detection under variable illumination is challenging due to the information loss caused by the limited dynamic range of conventional frame-based cameras. To address this issue, we introduce bio-inspired event cameras and propose a novel Structure-aware Fusion Network (SFNet) that extracts sharp and complete object structures from the event stream to compensate for the lost information in images through cross-modality fusion, enabling the network to obtain illumination-robust representations for traffic object detection. Specifically, to mitigate the sparsity or blurriness issues arising from diverse motion states of traffic objects in fixed-interval event sampling methods, we propose the Reliable Structure Generation Network (RSGNet) to generate Speed Invariant Frames (SIF), ensuring the integrity and sharpness of object structures. Next, we design a novel Adaptive Feature Complement Module (AFCM) which guides the adaptive fusion of two modality features to compensate for the information loss in the images by perceiving the global lightness distribution of the images, thereby generating illumination-robust representations. Finally, considering the lack of large-scale and high-quality annotations in the existing event-based object detection datasets, we build a DSEC-Det dataset, which consists of 53 sequences with 63,931 images and more than 208,000 labels for 8 classes. Extensive experimental results demonstrate that our proposed SFNet can overcome the perceptual boundaries of conventional cameras and outperform the frame-based method by 8.0% in mAP50 and 5.9% in mAP50:95. Our code and dataset will be available at https://github.com/YN-Yang/SFNet.

Rui Deng, Qian Wu, Yuke Li et al.

Optimizing video inference efficiency has become increasingly important with the growing demand for video analysis in various fields. Some existing methods achieve high efficiency by explicit discard of spatial or temporal information, which poses challenges in fast-changing and fine-grained scenarios. To address these issues, we propose an efficient video representation network with Differentiable Resolution Compression and Alignment mechanism, which compresses non-essential information in the early stage of the network to reduce computational costs while maintaining consistent temporal correlations. Specifically, we leverage a Differentiable Context-aware Compression Module to encode the saliency and non-saliency frame features, refining and updating the features into a high-low resolution video sequence. To process the new sequence, we introduce a new Resolution-Align Transformer Layer to capture global temporal correlations among frame features with different resolutions, while reducing spatial computation costs quadratically by utilizing fewer spatial tokens in low-resolution non-saliency frames. The entire network can be end-to-end optimized via the integration of the differentiable compression module. Experimental results show that our method achieves the best trade-off between efficiency and performance on near-duplicate video retrieval and competitive results on dynamic video classification compared to state-of-the-art methods. Code:https://github.com/dun-research/DRCA

Ke Xiong, Qian Wu, Wangjie Gan et al.

Few-shot Hierarchical Text Classification (few-shot HTC) is a challenging task that involves mapping texts to a predefined tree-structured label hierarchy under data-scarce conditions. While current approaches utilize structural constraints from the label hierarchy to maintain parent-child prediction consistency, they face a critical bottleneck, the difficulty in distinguishing semantically similar sibling classes due to insufficient domain knowledge. We introduce an innovative method named Sibling Contrastive Learning with Hierarchical Knowledge-aware Prompt Tuning for few-shot HTC tasks (SCHK-HTC). Our work enhances the model's perception of subtle differences between sibling classes at deeper levels, rather than just enforcing hierarchical rules. Specifically, we propose a novel framework featuring two core components: a hierarchical knowledge extraction module and a sibling contrastive learning mechanism. This design guides model to encode discriminative features at each hierarchy level, thus improving the separability of confusable classes. Our approach achieves superior performance across three benchmark datasets, surpassing existing state-of-the-art methods in most cases. Our code is available at https://github.com/happywinder/SCHK-HTC.

Yuke Li, Lianli Gao, Ji Zhang et al.

Exemplar-guided Image Editing (EIE) aims to modify a source image according to a visual reference. Existing approaches often require large-scale pre-training to learn relationships between the source and reference images, incurring high computational costs. As a training-free alternative, inversion techniques can be used to map the source image into a latent space for manipulation. However, our empirical study reveals that standard inversion is sub-optimal for EIE, leading to poor quality and inefficiency. To tackle this challenge, we introduce \textbf{Reversible Inversion ({ReInversion})} for effective and efficient EIE. Specifically, ReInversion operates as a two-stage denoising process, which is first conditioned on the source image and subsequently on the reference. Besides, we introduce a Mask-Guided Selective Denoising (MSD) strategy to constrain edits to target regions, preserving the structural consistency of the background. Both qualitative and quantitative comparisons demonstrate that our ReInversion method achieves state-of-the-art EIE performance with the lowest computational overhead.

Fengxian Ji, Jingpu Yang, Zirui Song et al.

Despite the rapid progress of large vision-language models (LVLMs), fine-grained, state-conditioned GUI interaction remains challenging. Current evaluations offer limited coverage, imprecise target-state definitions, and an overreliance on final-task success, obscuring where and why agents fail. To address this gap, we introduce \textbf{FineState-Bench}, a benchmark that evaluates whether an agent can correctly ground an instruction to the intended UI control and reach the exact target state. FineState-Bench comprises 2,209 instances across desktop, web, and mobile platforms, spanning four interaction families and 23 UI component types, with each instance explicitly specifying an exact target state for fine-grained state setting. We further propose \textit{FineState-Metrics}, a four-stage diagnostic pipeline with stage-wise success rates: Localization Success Rate (SR@Loc), Interaction Success Rate (SR@Int), Exact State Success Rate at Locate (ES-SR@Loc), and Exact State Success Rate at Interact (ES-SR@Int), and a plug-and-play \textit{Visual Diagnostic Assistant} (VDA) that generates a Description and a bounding-box Localization Hint to diagnose visual grounding reason via controlled w/ vs.\ w/o comparisons. On FineState-Bench, exact goal-state success remains low: ES-SR@Int peaks at 32.8\% on Web and 22.8\% on average across platforms. With VDA localization hints, Gemini-2.5-Flash gains +14.9 ES-SR@Int points, suggesting substantial headroom from improved visual grounding, yet overall accuracy is still insufficient for reliable fine-grained state-conditioned interaction \href{https://github.com/FengxianJi/FineState-Bench}{Github.}

Yuke Li, Weihang Liu, Cheng Zhang et al.

Feed-forward 3D Gaussian Splatting (3DGS) models offer fast single-pass reconstruction,but scaling them to match per-scene optimization quality is fundamentally hindered by the scarcity of large-scale 3D annotations.A practical compromise is predict-then-refine,where post-prediction optimization compensates for the limited capacity of the feed-forward network.However,standard feed-forward 3DGS is trained solely for zero-step rendering error,ignoring whether its output constitutes a good initialization for the downstream optimizer.We present ForeSplat,an optimization-aware training framework that equips feed-forward 3DGS models to produce initializations explicitly designed for rapid,effective refinement.By offloading part of the scene-modeling burden to the optimizer,ForeSplat substantially reduces the capacity pressure on the feed-forward model,making high-quality reconstruction feasible even with compact networks.At its core is MetaGrad,a lightweight multi-anchor meta-gradient training rule that bypasses costly higher-order differentiation through the 3DGS optimizer.MetaGrad unrolls a short inner-loop refinement trajectory,samples anchor states,and back-propagates aggregated first-order gradients to the prediction head as a surrogate optimization-aware signal.This fine-tuning adds no inference cost and enables high-quality reconstruction within seconds after a few refinement steps.We instantiate ForeSplat on diverse backbones,including AnySplat,Pi3X,and a distilled variant tailored for edge deployment.Across all tested architectures,a ForeSplat-trained initialization converges in fewer refinement steps and reaches a higher peak reconstruction quality than its vanilla counterpart,even fully converged.The framework consistently bridges the gap between amortized prediction and per-scene optimization,establishing a practical path toward lightweight,high-fidelity 3D reconstruction.

Kaifeng Wei, Teng Liu, Liang Dong et al.

Graph Anomaly Detection (GAD) is increasingly shifting to Generalist GAD (GGAD) for cross-domain "one-for-all" detection, but existing GGAD methods predominantly rely on the neighbor consistency principle, falling into the \textbf{Node-to-Neighbor Consistency Paradigm} for anomaly quantification. These methods suffer from complex training pipelines, heavy training data dependency, high computational costs, and unstable cross-domain generalization. To address these limitations, we propose NeighborDiv, a training-free generalist graph anomaly detection framework based on neighbor diversity. Departing from the dominant Node-to-Neighbor Consistency Paradigm, we shift the focus to the \textbf{Neighbor-to-Neighbor Diversity Paradigm}, and uncover that the internal structural dispersion of a node's neighbor set is a powerful, independently discriminative anomaly signal. We quantify neighbor diversity via the variance of inter-neighbor feature similarities, which captures how a node organizes its local graph environment, and operates independently of conventional node-to-neighbor consistency frameworks. Extensive experiments under two standard GGAD evaluation paradigms show NeighborDiv achieves state-of-the-art performance, with relative gains of 10.25% in average AUC and 17.78% in average AP over the second-best baseline under Single-Domain Independent Training (SDIT), and 6.89%/9.58% in AUC/AP under Unified Multi-Domain Training (UMDT), respectively. Notably, NeighborDiv yields zero performance volatility across all datasets, eliminating training-set dependency and establishing a lightweight and highly practical GGAD framework.

Xinyao Li, Yuke Li, Zhekai Du et al.

Large vision-language models (VLMs) like CLIP have demonstrated good zero-shot learning performance in the unsupervised domain adaptation task. Yet, most transfer approaches for VLMs focus on either the language or visual branches, overlooking the nuanced interplay between both modalities. In this work, we introduce a Unified Modality Separation (UniMoS) framework for unsupervised domain adaptation. Leveraging insights from modality gap studies, we craft a nimble modality separation network that distinctly disentangles CLIP's features into language-associated and vision-associated components. Our proposed Modality-Ensemble Training (MET) method fosters the exchange of modality-agnostic information while maintaining modality-specific nuances. We align features across domains using a modality discriminator. Comprehensive evaluations on three benchmarks reveal our approach sets a new state-of-the-art with minimal computational costs. Code: https://github.com/TL-UESTC/UniMoS

Rui Deng, Qian Wu, Yuke Li

In this paper, we introduce 3D-CSL, a compact pipeline for Near-Duplicate Video Retrieval (NDVR), and explore a novel self-supervised learning strategy for video similarity learning. Most previous methods only extract video spatial features from frames separately and then design kinds of complex mechanisms to learn the temporal correlations among frame features. However, parts of spatiotemporal dependencies have already been lost. To address this, our 3D-CSL extracts global spatiotemporal dependencies in videos end-to-end with a 3D transformer and find a good balance between efficiency and effectiveness by matching on clip-level. Furthermore, we propose a two-stage self-supervised similarity learning strategy to optimize the entire network. Firstly, we propose PredMAE to pretrain the 3D transformer with video prediction task; Secondly, ShotMix, a novel video-specific augmentation, and FCS loss, a novel triplet loss, are proposed further promote the similarity learning results. The experiments on FIVR-200K and CC_WEB_VIDEO demonstrate the superiority and reliability of our method, which achieves the state-of-the-art performance on clip-level NDVR.

Yuting Yang, Yuke Li, Binbin Du

The CTC-based automatic speech recognition (ASR) models without the external language model usually lack the capacity to model conditional dependencies and textual interactions. In this paper, we present a Gated Interlayer Collaboration (GIC) mechanism to improve the performance of CTC-based models, which introduces textual information into the model and thus relaxes the conditional independence assumption of CTC-based models. Specifically, we consider the weighted sum of token embeddings as the textual representation for each position, where the position-specific weights are the softmax probability distribution constructed via inter-layer auxiliary CTC losses. The textual representations are then fused with acoustic features by developing a gate unit. Experiments on AISHELL-1, TEDLIUM2, and AIDATATANG corpora show that the proposed method outperforms several strong baselines.

Qian Wu, Ruoxuan Cui, Yuke Li et al.

Action recognition in videos poses a challenge due to its high computational cost, especially for Joint Space-Time video transformers (Joint VT). Despite their effectiveness, the excessive number of tokens in such architectures significantly limits their efficiency. In this paper, we propose HaltingVT, an efficient video transformer adaptively removing redundant video patch tokens, which is primarily composed of a Joint VT and a Glimpser module. Specifically, HaltingVT applies data-adaptive token reduction at each layer, resulting in a significant reduction in the overall computational cost. Besides, the Glimpser module quickly removes redundant tokens in shallow transformer layers, which may even be misleading for video recognition tasks based on our observations. To further encourage HaltingVT to focus on the key motion-related information in videos, we design an effective Motion Loss during training. HaltingVT acquires video analysis capabilities and token halting compression strategies simultaneously in a unified training process, without requiring additional training procedures or sub-networks. On the Mini-Kinetics dataset, we achieved 75.0% top-1 ACC with 24.2 GFLOPs, as well as 67.2% top-1 ACC with an extremely low 9.9 GFLOPs. The code is available at https://github.com/dun-research/HaltingVT.

Jingxiao Yang, DaLin He, Miao Pan et al.

Recent advances in Large Vision-Language Models (LVLMs) have enabled training-free Fine-Grained Visual Recognition (FGVR). However, effectively exploiting LVLMs for FGVR remains challenging due to the inherent visual ambiguity of subordinate-level categories. Existing methods predominantly adopt either retrieval-oriented or reasoning-oriented paradigms to tackle this challenge, but both are constrained by two fundamental limitations:(1) They apply the same inference pipeline to all samples without accounting for uneven recognition difficulty, thereby leading to suboptimal accuracy and efficiency; (2) The lack of mechanisms to consolidate and reuse error-specific experience causes repeated failures on similar challenging cases. To address these limitations, we propose SARE, a Sample-wise Adaptive textbfREasoning framework for training-free FGVR. Specifically, SARE adopts a cascaded design that combines fast candidate retrieval with fine-grained reasoning, invoking the latter only when necessary. In the reasoning process, SARE incorporates a self-reflective experience mechanism that leverages past failures to provide transferable discriminative guidance during inference, without any parameter updates. Extensive experiments across 14 datasets substantiate that SARE achieves state-of-the-art performance while substantially reducing computational overhead.

Yuting Yang, Binbin Du, Yuke Li

The choice of modeling units is crucial for automatic speech recognition (ASR) tasks. In mandarin scenarios, the Chinese characters represent meaning but are not directly related to the pronunciation. Thus only considering the writing of Chinese characters as modeling units is insufficient to capture speech features. In this paper, we present a novel method involves with multi-level modeling units, which integrates multi-level information for mandarin speech recognition. Specifically, the encoder block considers syllables as modeling units and the decoder block deals with character-level modeling units. To facilitate the incremental conversion from syllable features to character features, we design an auxiliary task that applies cross-entropy (CE) loss to intermediate decoder layers. During inference, the input feature sequences are converted into syllable sequences by the encoder block and then converted into Chinese characters by the decoder block. Experiments on the widely used AISHELL-1 corpus demonstrate that our method achieves promising results with CER of 4.1%/4.6% and 4.6%/5.2%, using the Conformer and the Transformer backbones respectively.

Yujia Zheng, Fan Feng, Yuke Li et al.

Given a generalist model, learning a task-relevant specialist representation is fundamental for downstream applications. Identifiability, the asymptotic guarantee of recovering the ground-truth representation, is critical because it sets the ultimate limit of any model, even with infinite data and computation. We study this problem in a completely nonparametric setting, without relying on interventions, parametric forms, or structural constraints. We first prove that the structure between time steps and tasks is identifiable in a fully unsupervised manner, even when sequences lack strict temporal dependence and may exhibit disconnections, and task assignments can follow arbitrarily complex and interleaving structures. We then prove that, within each time step, the task-relevant latent representation can be disentangled from the irrelevant part under a simple sparsity regularization, without any additional information or parametric constraints. Together, these results establish a hierarchical foundation: task structure is identifiable across time steps, and task-relevant latent representations are identifiable within each step. To our knowledge, each result provides a first general nonparametric identifiability guarantee, and together they mark a step toward provably moving from generalist to specialist models.

Yuting Yang, Yuke Li, Binbin Du

The unified streaming and non-streaming speech recognition model has achieved great success due to its comprehensive capabilities. In this paper, we propose to improve the accuracy of the unified model by bridging the inherent representation gap between the streaming and non-streaming modes with a contrastive objective. Specifically, the top-layer hidden representation at the same frame of the streaming and non-streaming modes are regarded as a positive pair, encouraging the representation of the streaming mode close to its non-streaming counterpart. The multiple negative samples are randomly selected from the rest frames of the same sample under the non-streaming mode. Experimental results demonstrate that the proposed method achieves consistent improvements toward the unified model in both streaming and non-streaming modes. Our method achieves CER of 4.66% in the streaming mode and CER of 4.31% in the non-streaming mode, which sets a new state-of-the-art on the AISHELL-1 benchmark.

Fengxian Ji, Jingpu Yang, Zirui Song et al.

With the rapid advancement of generative artificial intelligence technology, Graphical User Interface (GUI) agents have demonstrated tremendous potential for autonomously managing daily tasks through natural language instructions. However, current evaluation frameworks for GUI agents suffer from fundamental flaws: existing benchmarks overly focus on coarse-grained task completion while neglecting fine-grained control capabilities crucial for real-world applications. To address this, we introduce FineState-Bench, the first evaluation and diagnostic standard for fine-grained GUI proxy operations, designed to quantify fine-grained control. This multi-platform (desktop, Web, mobile) framework includes 2257 task benchmarks in four components and uses a four-phase indicator for comprehensive perception-to-control assessment. To analyze perception and positioning for refined operations, we developed the plug-and-play Visual Diagnostic Assistant (VDA), enabling the first quantitative decoupling analysis of these capabilities. Experimental results on our benchmark show that the most advanced models achieve only 32.8% fine-grained interaction accuracy. Using our VDA in controlled experiments, quantifying the impact of visual capabilities, we showed that ideal visual localization boosts Gemini-2.5-Flash's success rate by 14.9\%. Our diagnostic framework confirms for the first time that the primary bottleneck for current GUI proxies is basic visual positioning capability.All resources are fully open-source. github: https://github.com/AnonymousThewarehouse/FineState-Bench huggingface: https://huggingface.co/datasets/Willtime2006/Static-FineBench

Lifeng Zhou, Yuke Li, Rui Deng et al.

The success of speech-image retrieval relies on establishing an effective alignment between speech and image. Existing methods often model cross-modal interaction through simple cosine similarity of the global feature of each modality, which fall short in capturing fine-grained details within modalities. To address this issue, we introduce an effective framework and a novel learning task named cross-modal denoising (CMD) to enhance cross-modal interaction to achieve finer-level cross-modal alignment. Specifically, CMD is a denoising task designed to reconstruct semantic features from noisy features within one modality by interacting features from another modality. Notably, CMD operates exclusively during model training and can be removed during inference without adding extra inference time. The experimental results demonstrate that our framework outperforms the state-of-the-art method by 2.0% in mean R@1 on the Flickr8k dataset and by 1.7% in mean R@1 on the SpokenCOCO dataset for the speech-image retrieval tasks, respectively. These experimental results validate the efficiency and effectiveness of our framework.

Wenjun Liu, Qian Wu, Yifeng Hu et al.

We introduce SELECT (Scene tExt Label Errors deteCTion), a novel approach that leverages multi-modal training to detect label errors in real-world scene text datasets. Utilizing an image-text encoder and a character-level tokenizer, SELECT addresses the issues of variable-length sequence labels, label sequence misalignment, and character-level errors, outperforming existing methods in accuracy and practical utility. In addition, we introduce Similarity-based Sequence Label Corruption (SSLC), a process that intentionally introduces errors into the training labels to mimic real-world error scenarios during training. SSLC not only can cause a change in the sequence length but also takes into account the visual similarity between characters during corruption. Our method is the first to detect label errors in real-world scene text datasets successfully accounting for variable-length labels. Experimental results demonstrate the effectiveness of SELECT in detecting label errors and improving STR accuracy on real-world text datasets, showcasing its practical utility.

Lifeng Zhou, Yuke Li

In this paper, we propose a novel framework for speech-image retrieval. We utilize speech-image contrastive (SIC) learning tasks to align speech and image representations at a coarse level and speech-image matching (SIM) learning tasks to further refine the fine-grained cross-modal alignment. SIC and SIM learning tasks are jointly trained in a unified manner. To optimize the learning process, we utilize an embedding queue that facilitates efficient sampling of high-quality and diverse negative representations during SIC learning. Additionally, it enhances the learning of SIM tasks by effectively mining hard negatives based on contrastive similarities calculated in SIC tasks. To further optimize learning under noisy supervision, we incorporate momentum distillation into the training process. Experimental results show that our framework outperforms the state-of-the-art method by more than 4% in R@1 on two benchmark datasets for the speech-image retrieval tasks. Moreover, as observed in zero-shot experiments, our framework demonstrates excellent generalization capabilities.

Runqi Wang, Caoyuan Ma, Guopeng Li et al.

Text to Motion aims to generate human motions from texts. Existing settings rely on limited Action Texts that include action labels, which limits flexibility and practicability in scenarios difficult to describe directly. This paper extends limited Action Texts to arbitrary ones. Scene texts without explicit action labels can enhance the practicality of models in complex and diverse industries such as virtual human interaction, robot behavior generation, and film production, while also supporting the exploration of potential implicit behavior patterns. However, newly introduced Scene Texts may yield multiple reasonable output results, causing significant challenges in existing data, framework, and evaluation. To address this practical issue, we first create a new dataset HUMANML3D++ by extending texts of the largest existing dataset HUMANML3D. Secondly, we propose a simple yet effective framework that extracts action instructions from arbitrary texts and subsequently generates motions. Furthermore, we also benchmark this new setting with multi-solution metrics to address the inadequacies of existing single-solution metrics. Extensive experiments indicate that Text to Motion in this realistic setting is challenging, fostering new research in this practical direction.

Guodong Ma, Wenxuan Wang, Lifeng Zhou et al.

Recently, the Mixture of Expert (MoE) architecture, such as LR-MoE, is often used to alleviate the impact of language confusion on the multilingual ASR (MASR) task. However, it still faces language confusion issues, especially in mismatched domain scenarios. In this paper, we decouple language confusion in LR-MoE into confusion in self-attention and router. To alleviate the language confusion in self-attention, based on LR-MoE, we propose to apply attention-MoE architecture for MASR. In our new architecture, MoE is utilized not only on feed-forward network (FFN) but also on self-attention. In addition, to improve the robustness of the LID-based router on language confusion, we propose expert pruning and router augmentation methods. Combining the above, we get the boosted language-routing MoE (BLR-MoE) architecture. We verify the effectiveness of the proposed BLR-MoE in a 10,000-hour MASR dataset.

Yuke Li, Guangyi Chen, Ben Abramowitz et al.

Few-shot action recognition aims at quickly adapting a pre-trained model to the novel data with a distribution shift using only a limited number of samples. Key challenges include how to identify and leverage the transferable knowledge learned by the pre-trained model. We therefore propose CDTD, or Causal Domain-Invariant Temporal Dynamics for knowledge transfer. To identify the temporally invariant and variant representations, we employ the causal representation learning methods for unsupervised pertaining, and then tune the classifier with supervisions in next stage. Specifically, we assume the domain information can be well estimated and the pre-trained image decoder and transition models can be well transferred. During adaptation, we fix the transferable temporal dynamics and update the image encoder and domain estimator. The efficacy of our approach is revealed by the superior accuracy of CDTD over leading alternatives across standard few-shot action recognition datasets.

Junbo Wang, Liangyu Fu, Yuke Li et al.

Micro-expression recognition can obtain the real emotion of the individual at the current moment. Although deep learning-based methods, especially Transformer-based methods, have achieved impressive results, these methods have high computational complexity due to the large number of tokens in the multi-head self-attention. In addition, the existing micro-expression datasets are small-scale, which makes it difficult for Transformer-based models to learn effective micro-expression representations. Therefore, we propose a novel Efficient Patch tokenization, Integration and Representation framework (EPIR), which can balance high recognition performance and low computational complexity. Specifically, we first propose a dual norm shifted tokenization (DNSPT) module to learn the spatial relationship between neighboring pixels in the face region, which is implemented by a refined spatial transformation and dual norm projection. Then, we propose a token integration module to integrate partial tokens among multiple cascaded Transformer blocks, thereby reducing the number of tokens without information loss. Furthermore, we design a discriminative token extractor, which first improves the attention in the Transformer block to reduce the unnecessary focus of the attention calculation on self-tokens, and uses the dynamic token selection module (DTSM) to select key tokens, thereby capturing more discriminative micro-expression representations. We conduct extensive experiments on four popular public datasets (i.e., CASME II, SAMM, SMIC, and CAS(ME)3. The experimental results show that our method achieves significant performance gains over the state-of-the-art methods, such as 9.6% improvement on the CAS(ME)$^3$ dataset in terms of UF1 and 4.58% improvement on the SMIC dataset in terms of UAR metric.

Junbo Wang, Liangyu Fu, Yuke Li et al.

Current video captioning methods usually use an encoder-decoder structure to generate text autoregressively. However, autoregressive methods have inherent limitations such as slow generation speed and large cumulative error. Furthermore, the few non-autoregressive counterparts suffer from deficiencies in generation quality due to the lack of sufficient multimodal interaction modeling. Therefore, we propose a non-autoregressive framework based on Diffusion model for Video Captioning (DiffVC) to address these issues. Its parallel decoding can effectively solve the problems of generation speed and cumulative error. At the same time, our proposed discriminative conditional Diffusion Model can generate higher-quality textual descriptions. Specifically, we first encode the video into a visual representation. During training, Gaussian noise is added to the textual representation of the ground-truth caption. Then, a new textual representation is generated via the discriminative denoiser with the visual representation as a conditional constraint. Finally, we input the new textual representation into a non-autoregressive language model to generate captions. During inference, we directly sample noise from the Gaussian distribution for generation. Experiments on MSVD, MSR-VTT, and VATEX show that our method can outperform previous non-autoregressive methods and achieve comparable performance to autoregressive methods, e.g., it achieved a maximum improvement of 9.9 on the CIDEr and improvement of 2.6 on the B@4, while having faster generation speed. The source code will be available soon.

Weihang Liu, Yuhui Zhong, Yuke Li et al.

Accurate and efficient modeling of large-scale urban scenes is critical for applications such as AR navigation, UAV based inspection, and smart city digital twins. While aerial imagery offers broad coverage and complements limitations of ground-based data, reconstructing city-scale environments from such views remains challenging due to occlusions, incomplete geometry, and high memory demands. Recent advances like 3D Gaussian Splatting (3DGS) improve scalability and visual quality but remain limited by dense primitive usage, long training times, and poor suit ability for edge devices. We propose CityGo, a hybrid framework that combines textured proxy geometry with residual and surrounding 3D Gaussians for lightweight, photorealistic rendering of urban scenes from aerial perspectives. Our approach first extracts compact building proxy meshes from MVS point clouds, then uses zero order SH Gaussians to generate occlusion-free textures via image-based rendering and back-projection. To capture high-frequency details, we introduce residual Gaussians placed based on proxy-photo discrepancies and guided by depth priors. Broader urban context is represented by surrounding Gaussians, with importance-aware downsampling applied to non-critical regions to reduce redundancy. A tailored optimization strategy jointly refines proxy textures and Gaussian parameters, enabling real-time rendering of complex urban scenes on mobile GPUs with significantly reduced training and memory requirements. Extensive experiments on real-world aerial datasets demonstrate that our hybrid representation significantly reduces training time, achieving on average 1.4x speedup, while delivering comparable visual fidelity to pure 3D Gaussian Splatting approaches. Furthermore, CityGo enables real-time rendering of large-scale urban scenes on mobile consumer GPUs, with substantially reduced memory usage and energy consumption.

Zhengwei Yang, Yuke Li, Qiang Sun et al.

Most existing studies on few-shot learning focus on unimodal settings, where models are trained to generalize to unseen data using a limited amount of labeled examples from a single modality. However, real-world data are inherently multi-modal, and such unimodal approaches limit the practical applications of few-shot learning. To bridge this gap, this paper introduces the Cross-modal Few-Shot Learning (CFSL) task, which aims to recognize instances across multiple modalities while relying on scarce labeled data. This task presents unique challenges compared to classical few-shot learning arising from the distinct visual attributes and structural disparities inherent to each modality. To tackle these challenges, we propose a Generative Transfer Learning (GTL) framework by simulating how humans abstract and generalize concepts. Specifically, the GTL jointly estimates the latent shared concept across modalities and the in-modality disturbance through a generative structure. Establishing the relationship between latent concepts and visual content among abundant unimodal data enables GTL to effectively transfer knowledge from unimodal to novel multimodal data, as humans did. Comprehensive experiments demonstrate that the GTL achieves state-of-the-art performance across seven multi-modal datasets across RGB-Sketch, RGB-Infrared, and RGB-Depth.

Yuke Li, Yujia Zheng, Tianyi Xiong et al.

Catastrophic interference, also known as catastrophic forgetting, is a fundamental challenge in machine learning, where a trained learning model progressively loses performance on previously learned tasks when adapting to new ones. In this paper, we aim to better understand and model the catastrophic interference problem from a latent representation learning point of view, and propose a novel theoretical framework that formulates catastrophic interference as an identification problem. Our analysis demonstrates that the forgetting phenomenon can be quantified by the distance between partial-task aware (PTA) and all-task aware (ATA) setups. Building upon recent advances in identifiability theory, we prove that this distance can be minimized through identification of shared latent variables between these setups. When learning, we propose our method \ourmeos with two-stage training strategy: First, we employ maximum likelihood estimation to learn the latent representations from both PTA and ATA configurations. Subsequently, we optimize the KL divergence to identify and learn the shared latent variables. Through theoretical guarantee and empirical validations, we establish that identifying and learning these shared representations can effectively mitigate catastrophic interference in machine learning systems. Our approach provides both theoretical guarantees and practical performance improvements across both synthetic and benchmark datasets.

Weihang Liu, Yuke Li, Yuxuan Li et al.

Recent advances in 3D Gaussian Splatting (3DGS) have demonstrated remarkable rendering fidelity and efficiency. However, these methods still rely on computationally expensive sequential alpha-blending operations, resulting in significant overhead, particularly on resource-constrained platforms. In this paper, we propose Duplex-GS, a dual-hierarchy framework that integrates proxy Gaussian representations with order-independent rendering techniques to achieve photorealistic results while sustaining real-time performance. To mitigate the overhead caused by view-adaptive radix sort, we introduce cell proxies for local Gaussians management and propose cell search rasterization for further acceleration. By seamlessly combining our framework with Order-Independent Transparency (OIT), we develop a physically inspired weighted sum rendering technique that simultaneously eliminates "popping" and "transparency" artifacts, yielding substantial improvements in both accuracy and efficiency. Extensive experiments on a variety of real-world datasets demonstrate the robustness of our method across diverse scenarios, including multi-scale training views and large-scale environments. Our results validate the advantages of the OIT rendering paradigm in Gaussian Splatting, achieving high-quality rendering with an impressive 1.5 to 4 speedup over existing OIT based Gaussian Splatting approaches and 52.2% to 86.9% reduction of the radix sort overhead without quality degradation.

Junfei Shi, Yuke Li, Mengmeng Nie et al.

Deep learning has been extensively utilized for PolSAR image classification. However, most existing methods transform the polarimetric covariance matrix into a real- or complex-valued vector to comply with standard deep learning frameworks in Euclidean space. This approach overlooks the inherent structure of the covariance matrix, which is a complex Hermitian positive definite (HPD) matrix residing in the Riemannian manifold. Vectorization disrupts the matrix structure and misrepresents its geometric properties. To mitigate this drawback, we propose HPDNet, a novel framework that directly processes HPD matrices on the Riemannian manifold. The HPDnet fully considers the complex phase information by decomposing a complex HPD matrix into the real- and imaginarymatrices. The proposed HPDnet consists of several HPD mapping layers and rectifying layers, which can preserve the geometric structure of the data and transform them into a more separable manifold representation. Subsequently, a complex LogEig layer is developed to project the manifold data into a tangent space, ensuring that conventional Euclidean-based deep learning networks can be applied to further extract contextual features for classification. Furthermore, to optimize computational efficiency, we design a fast eigenvalue decomposition method for parallelized matrix processing. Experiments conducted on three real-world PolSAR datasets demonstrate that the proposed method outperforms state-of-the-art approaches, especially in heterogeneous regions.

Yuke Li, Lixiong Chen, Guangyi Chen et al.

In order to predict a pedestrian's trajectory in a crowd accurately, one has to take into account her/his underlying socio-temporal interactions with other pedestrians consistently. Unlike existing work that represents the relevant information separately, partially, or implicitly, we propose a complete representation for it to be fully and explicitly captured and analyzed. In particular, we introduce a Directed Acyclic Graph-based structure, which we term Socio-Temporal Graph (STG), to explicitly capture pair-wise socio-temporal interactions among a group of people across both space and time. Our model is built on a time-varying generative process, whose latent variables determine the structure of the STGs. We design an attention-based model named STGformer that affords an end-to-end pipeline to learn the structure of the STGs for trajectory prediction. Our solution achieves overall state-of-the-art prediction accuracy in two large-scale benchmark datasets. Our analysis shows that a person's past trajectory is critical for predicting another person's future path. Our model learns this relationship with a strong notion of socio-temporal localities. Statistics show that utilizing this information explicitly for prediction yields a noticeable performance gain with respect to the trajectory-only approaches.

Yuting Yang, Binbin Du, Yingxin Zhang et al.

This paper introduces the system submitted by the Yidun NISP team to the video keyword wakeup challenge. We propose a mandarin keyword spotting system (KWS) with several novel and effective improvements, including a big backbone (B) model, a keyword biasing (B) mechanism and the introduction of syllable modeling units (S). By considering this, we term the total system BBS-KWS as an abbreviation. The BBS-KWS system consists of an end-to-end automatic speech recognition (ASR) module and a KWS module. The ASR module converts speech features to text representations, which applies a big backbone network to the acoustic model and takes syllable modeling units into consideration as well. In addition, the keyword biasing mechanism is used to improve the recall rate of keywords in the ASR inference stage. The KWS module applies multiple criteria to determine the absence or presence of the keywords, such as multi-stage matching, fuzzy matching, and connectionist temporal classification (CTC) prefix score. To further improve our system, we conduct semi-supervised learning on the CN-Celeb dataset for better generalization. In the VKW task, the BBS-KWS system achieves significant gains over the baseline and won the first place in two tracks.

Yue Wang, Xu Jia, Lu Zhang et al.

RGB-D saliency detection integrates information from both RGB images and depth maps to improve prediction of salient regions under challenging conditions. The key to RGB-D saliency detection is to fully mine and fuse information at multiple scales across the two modalities. Previous approaches tend to apply the multi-scale and multi-modal fusion separately via local operations, which fails to capture long-range dependencies. Here we propose a transformer-based network to address this issue. Our proposed architecture is composed of two modules: a transformer-based within-modality feature enhancement module (TWFEM) and a transformer-based feature fusion module (TFFM). TFFM conducts a sufficient feature fusion by integrating features from multiple scales and two modalities over all positions simultaneously. TWFEM enhances feature on each scale by selecting and integrating complementary information from other scales within the same modality before TFFM. We show that transformer is a uniform operation which presents great efficacy in both feature fusion and feature enhancement, and simplifies the model design. Extensive experimental results on six benchmark datasets demonstrate that our proposed network performs favorably against state-of-the-art RGB-D saliency detection methods.

Bernice Rogowitz, Laura J. Perovich, Yuke Li et al.

It is human to want to touch artworks, to feel their surface curvature and texture, their shapes and structures, and to feel the hand of the artist. Museum guards need to be constantly vigilant to protect art objects from adoring and exploring touches by visitors. This paper introduces a novel technique for capturing where and how art objects are touched. In this method, the users' touch either adds, or subtracts, microscopic fluorescent particles from a three-dimensional art object. Viewing the object under ultraviolet light reveals their touch traces and gestures. We present human touch behavior for a three-dimensional stylized landscape, and for two abstract and two representational art objects. We also present the results of video recordings of real-time behavior and user interviews. The resulting data show the kinds of touches, and where they are directed, and also reveal important individual differences. We feel this method opens the door to studying art perception through touch, and also enables new kinds of studies into touch behavior in other applications, including visualization, embodied cognition, and design.

Yue Wang, Yuke Li, James H. Elder et al.

Depth information available from an RGB-D camera can be useful in segmenting salient objects when figure/ground cues from RGB channels are weak. This has motivated the development of several RGB-D saliency datasets and algorithms that use all four channels of the RGB-D data for both training and inference. Unfortunately, existing RGB-D saliency datasets are small, which may lead to overfitting and limited generalization for diverse scenarios. Here we propose a semi-supervised system for RGB-D saliency detection that can be trained on smaller RGB-D saliency datasets without saliency ground truth, while also make effective joint use of a large RGB saliency dataset with saliency ground truth together. To generalize our method on RGB-D saliency datasets, a novel prediction-guided cross-refinement module which jointly estimates both saliency and depth by mutual refinement between two respective tasks, and an adversarial learning approach are employed. Critically, our system does not require saliency ground-truth for the RGB-D datasets, which saves the massive human labor for hand labeling, and does not require the depth data for inference, allowing the method to be used for the much broader range of applications where only RGB data are available. Evaluation on seven RGB-D datasets demonstrates that even without saliency ground truth for RGB-D datasets and using only the RGB data of RGB-D datasets at inference, our semi-supervised system performs favorable against state-of-the-art fully-supervised RGB-D saliency detection methods that use saliency ground truth for RGB-D datasets at training and depth data at inference on two largest testing datasets. Our approach also achieves comparable results on other popular RGB-D saliency benchmarks.

Yue Wang, Yuke Li, James H. Elder et al.

Semantic segmentation is an important sub-task for many applications, but pixel-level ground truth labeling is costly and there is a tendency to overfit the training data, limiting generalization. Unsupervised domain adaptation can potentially address these problems, allowing systems trained on labelled datasets from one or more source domains (including less expensive synthetic domains) to be adapted to novel target domains. The conventional approach is to automatically align the representational distributions of source and target domains. One limitation of this approach is that it tends to disadvantage lower probability classes. We address this problem by introducing a Class-Conditional Domain Adaptation method (CCDA). It includes a class-conditional multi-scale discriminator and the class-conditional loss. This novel CCDA method encourages the network to shift the domain in a class-conditional manner, and it equalizes loss over classes. We evaluate our CCDA method on two transfer tasks and demonstrate performance comparable to state-of-the-art methods.