Yunheng Li

Chenxu Wang, Yuxuan Li, Yunheng Li et al.

Existing language-image pre-training for remote sensing object detection is constrained by Monolithic Label Learning, which relies on exhaustively enumerating open-set categories via black-box data to acquire fine-grained representations, creating a dependency incompatible with the domain's inherent data scarcity. To transcend this bottleneck, we propose SLIP-RS, establishing a Structured-Attribute Decoupling Paradigm that maps the open-ended category space into a finite, physically meaningful attribute space, unlocking fine-grained discriminability via explicit structural logic. This paradigm is realized via two technical pillars: (1) Structured-Attribute Contrastive Learning, which enforces the learning of decoupled intrinsic visual logic via combinatorial attribute augmentation; and (2) Conformal Attribute Reliability Engine, which leverages conformal prediction theory to rigorously distill high-fidelity supervision from noisy sources, yielding RS-Attribute-15M, the largest dataset with over 15 million attribute annotations. Extensive experiments demonstrate that SLIP-RS establishes unprecedented performance in fine-grained detection and cross-domain generalization, validating structured attributes as a vital foundation for remote sensing. Code: https://github.com/facias914/SLIP-RS.

Jinrong Zhang, Wujun Wen, Shenglan Liu et al.

The streaming temporal action segmentation (STAS) task, a supplementary task of temporal action segmentation (TAS), has not received adequate attention in the field of video understanding. Existing TAS methods are constrained to offline scenarios due to their heavy reliance on multimodal features and complete contextual information. The STAS task requires the model to classify each frame of the entire untrimmed video sequence clip by clip in time, thereby extending the applicability of TAS methods to online scenarios. However, directly applying existing TAS methods to SATS tasks results in significantly poor segmentation outcomes. In this paper, we thoroughly analyze the fundamental differences between STAS tasks and TAS tasks, attributing the severe performance degradation when transferring models to model bias and optimization dilemmas. We introduce an end-to-end streaming video temporal action segmentation model with reinforcement learning (SVTAS-RL). The end-to-end modeling method mitigates the modeling bias introduced by the change in task nature and enhances the feasibility of online solutions. Reinforcement learning is utilized to alleviate the optimization dilemma. Through extensive experiments, the SVTAS-RL model significantly outperforms existing STAS models and achieves competitive performance to the state-of-the-art TAS model on multiple datasets under the same evaluation criteria, demonstrating notable advantages on the ultra-long video dataset EGTEA. Code is available at https://github.com/Thinksky5124/SVTAS.

Yunheng Li, Hangyi Kuang, Hengrui Zhang et al.

Multimodal Chain-of-Thought (CoT) reasoning requires large vision-language models to construct reasoning trajectories that interleave perceptual grounding with multi-step inference. However, existing Reinforcement Learning with Verifiable Rewards (RLVR) methods typically optimize reasoning at a coarse granularity, treating CoT uniformly without distinguishing their varying degrees of visual grounding. In this work, we conduct a token-level analysis of multimodal reasoning trajectories and show that successful reasoning is characterized by structured token dynamics reflecting both perceptual grounding and exploratory inference. Building upon this analysis, we propose Perception-Exploration Policy Optimization (PEPO), which derives a perception prior from hidden state similarity and integrates it with token entropy through a smooth gating mechanism to produce token-level advantages. PEPO integrates seamlessly with existing RLVR frameworks such as GRPO and DAPO, requiring neither additional supervision nor auxiliary branches. Extensive experiments across diverse multimodal benchmarks demonstrate consistent and robust improvements over strong RL baselines, spanning geometry reasoning, visual grounding, visual puzzle solving, and few-shot classification, while maintaining stable training dynamics. Code: https://github.com/xzxxntxdy/PEPO

Yunheng Li, Hengrui Zhang, Meng-Hao Guo et al.

Universal video understanding requires modeling fine-grained visual and audio information over time in diverse real-world scenarios. However, the performance of existing models is primarily constrained by video-instruction data that represents complex audiovisual content as single, incomplete descriptions, lacking fine-grained organization and reliable annotation. To address this, we introduce: (i) ASID-1M, an open-source collection of one million structured, fine-grained audiovisual instruction annotations with single- and multi-attribute supervision; (ii) ASID-Verify, a scalable data curation pipeline for annotation, with automatic verification and refinement that enforces semantic and temporal consistency between descriptions and the corresponding audiovisual content; and (iii) ASID-Captioner, a video understanding model trained via Supervised Fine-Tuning (SFT) on the ASID-1M. Experiments across seven benchmarks covering audiovisual captioning, attribute-wise captioning, caption-based QA, and caption-based temporal grounding show that ASID-Captioner improves fine-grained caption quality while reducing hallucinations and improving instruction following. It achieves state-of-the-art performance among open-source models and is competitive with Gemini-3-Pro.

Yuxuan Li, Yuming Chen, Yunheng Li et al.

Heterogeneous multi-modal remote sensing object detection aims to accurately detect objects from diverse sensors (e.g., RGB, SAR, Infrared). Existing approaches largely adopt a late alignment paradigm, in which modality alignment and task-specific optimization are entangled during downstream fine-tuning. This tight coupling complicates optimization and often results in unstable training and suboptimal generalization. To address these limitations, we propose BabelRS, a unified language-pivoted pretraining framework that explicitly decouples modality alignment from downstream task learning. BabelRS comprises two key components: Concept-Shared Instruction Aligning (CSIA) and Layerwise Visual-Semantic Annealing (LVSA). CSIA aligns each sensor modality to a shared set of linguistic concepts, using language as a semantic pivot to bridge heterogeneous visual representations. To further mitigate the granularity mismatch between high-level language representations and dense detection objectives, LVSA progressively aggregates multi-scale visual features to provide fine-grained semantic guidance. Extensive experiments demonstrate that BabelRS stabilizes training and consistently outperforms state-of-the-art methods without bells and whistles. Code: https://github.com/zcablii/SM3Det.

Yuxuan Li, Xiang Li, Yunheng Li et al.

With the rapid advancement of remote sensing technology, high-resolution multi-modal imagery is now more widely accessible. Conventional Object detection models are trained on a single dataset, often restricted to a specific imaging modality and annotation format. However, such an approach overlooks the valuable shared knowledge across multi-modalities and limits the model's applicability in more versatile scenarios. This paper introduces a new task called Multi-Modal Datasets and Multi-Task Object Detection (M2Det) for remote sensing, designed to accurately detect horizontal or oriented objects from any sensor modality. This task poses challenges due to 1) the trade-offs involved in managing multi-modal modelling and 2) the complexities of multi-task optimization. To address these, we establish a benchmark dataset and propose a unified model, SM3Det (Single Model for Multi-Modal datasets and Multi-Task object Detection). SM3Det leverages a grid-level sparse MoE backbone to enable joint knowledge learning while preserving distinct feature representations for different modalities. Furthermore, it integrates a consistency and synchronization optimization strategy using dynamic learning rate adjustment, allowing it to effectively handle varying levels of learning difficulty across modalities and tasks. Extensive experiments demonstrate SM3Det's effectiveness and generalizability, consistently outperforming specialized models on individual datasets. The code is available at https://github.com/zcablii/SM3Det.

Xuying Zhang, Bo-Wen Yin, Yuming Chen et al.

Recent progress in the text-driven 3D stylization of a single object has been considerably promoted by CLIP-based methods. However, the stylization of multi-object 3D scenes is still impeded in that the image-text pairs used for pre-training CLIP mostly consist of an object. Meanwhile, the local details of multiple objects may be susceptible to omission due to the existing supervision manner primarily relying on coarse-grained contrast of image-text pairs. To overcome these challenges, we present a novel framework, dubbed TeMO, to parse multi-object 3D scenes and edit their styles under the contrast supervision at multiple levels. We first propose a Decoupled Graph Attention (DGA) module to distinguishably reinforce the features of 3D surface points. Particularly, a cross-modal graph is constructed to align the object points accurately and noun phrases decoupled from the 3D mesh and textual description. Then, we develop a Cross-Grained Contrast (CGC) supervision system, where a fine-grained loss between the words in the textual description and the randomly rendered images are constructed to complement the coarse-grained loss. Extensive experiments show that our method can synthesize high-quality stylized content and outperform the existing methods over a wide range of multi-object 3D meshes. Our code and results will be made publicly available

Wujun Wen, Yunheng Li, Zhuben Dong et al.

Temporal action segmentation (TAS) is a critical step toward long-term video understanding. Recent studies follow a pattern that builds models based on features instead of raw video picture information. However, we claim those models are trained complicatedly and limit application scenarios. It is hard for them to segment human actions of video in real time because they must work after the full video features are extracted. As the real-time action segmentation task is different from TAS task, we define it as streaming video real-time temporal action segmentation (SVTAS) task. In this paper, we propose a real-time end-to-end multi-modality model for SVTAS task. More specifically, under the circumstances that we cannot get any future information, we segment the current human action of streaming video chunk in real time. Furthermore, the model we propose combines the last steaming video chunk feature extracted by language model with the current image feature extracted by image model to improve the quantity of real-time temporal action segmentation. To the best of our knowledge, it is the first multi-modality real-time temporal action segmentation model. Under the same evaluation criteria as full video temporal action segmentation, our model segments human action in real time with less than 40% of state-of-the-art model computation and achieves 90% of the accuracy of the full video state-of-the-art model.

Yunheng Li, Yuxuan Li, Quansheng Zeng et al.

Pre-trained vision-language models (VLMs), such as CLIP, have demonstrated impressive zero-shot recognition capability, but still underperform in dense prediction tasks. Self-distillation recently is emerging as a promising approach for fine-tuning VLMs to better adapt to local regions without requiring extensive annotations. However, previous state-of-the-art approaches often suffer from significant `foreground bias', where models tend to wrongly identify background regions as foreground objects. To alleviate this issue, we propose DenseVLM, a framework designed to learn unbiased region-language alignment from powerful pre-trained VLM representations. To alleviate this issue, we propose DenseVLM, a framework designed to learn unbiased region-language alignment from powerful pre-trained VLM representations. DenseVLM leverages the pre-trained VLM to retrieve categories for unlabeled regions and then decouples the interference between foreground and background features. We show that DenseVLM can directly replace the original VLM in open-vocabulary object detection and image segmentation methods, leading to notable performance improvements. Furthermore, it exhibits promising zero-shot scalability when training on more extensive and diverse datasets. Our code is available at https://github.com/HVision-NKU/DenseVLM.

Yunheng Li, Jing Cheng, Shaoyong Jia et al.

This paper introduces TempSamp-R1, a new reinforcement fine-tuning framework designed to improve the effectiveness of adapting multimodal large language models (MLLMs) to video temporal grounding tasks. We reveal that existing reinforcement learning methods, such as Group Relative Policy Optimization (GRPO), rely on on-policy sampling for policy updates. However, in tasks with large temporal search spaces, this strategy becomes both inefficient and limited in performance, as it often fails to identify temporally accurate solutions. To address this limitation, TempSamp-R1 leverages ground-truth annotations as off-policy supervision to provide temporally precise guidance, effectively compensating for the sparsity and misalignment in on-policy solutions. To further stabilize training and reduce variance in reward-based updates, TempSamp-R1 provides a non-linear soft advantage computation method that dynamically reshapes the reward feedback via an asymmetric transformation. By employing a hybrid Chain-of-Thought (CoT) training paradigm, TempSamp-R1 optimizes a single unified model to support both CoT and non-CoT inference modes, enabling efficient handling of queries with varying reasoning complexity. Experimental results demonstrate that TempSamp-R1 outperforms GRPO-based baselines, establishing new state-of-the-art performance on benchmark datasets: Charades-STA (R1@0.7: 52.9%, +2.7%), ActivityNet Captions (R1@0.5: 56.0%, +5.3%), and QVHighlights (mAP: 30.0%, +3.0%). Moreover, TempSamp-R1 shows robust few-shot generalization capabilities under limited data. Code: https://github.com/HVision-NKU/TempSamp-R1

Quan-Sheng Zeng, Yunheng Li, Daquan Zhou et al.

Open-vocabulary image segmentation has been advanced through the synergy between mask generators and vision-language models like Contrastive Language-Image Pre-training (CLIP). Previous approaches focus on generating masks while aligning mask features with text embeddings during training. In this paper, we observe that relying on generated low-quality masks can weaken the alignment of vision and language in regional representations. This motivates us to present a new fine-tuning framework, named MaskCLIP++, which uses ground-truth masks instead of generated masks to enhance the mask classification capability of CLIP. Due to the limited diversity of image segmentation datasets with mask annotations, we propose incorporating a consistency alignment principle during fine-tuning, which alleviates categorical bias toward the fine-tuning dataset. After low-cost fine-tuning, MaskCLIP++ significantly improves the mask classification performance on multi-domain datasets. Combining with the mask generator in previous state-of-the-art mask-based open vocabulary segmentation methods, we achieve performance improvements of +1.7, +2.3, +2.1, +3.1, and +0.3 mIoU on the A-847, PC-459, A-150, PC-59, and PAS-20 datasets, respectively. Code is avaliable at https://github.com/HVision-NKU/MaskCLIPpp .

Yunheng Li, ZhongYu Li, Quansheng Zeng et al.

Pre-trained vision-language models, e.g., CLIP, have been successfully applied to zero-shot semantic segmentation. Existing CLIP-based approaches primarily utilize visual features from the last layer to align with text embeddings, while they neglect the crucial information in intermediate layers that contain rich object details. However, we find that directly aggregating the multi-level visual features weakens the zero-shot ability for novel classes. The large differences between the visual features from different layers make these features hard to align well with the text embeddings. We resolve this problem by introducing a series of independent decoders to align the multi-level visual features with the text embeddings in a cascaded way, forming a novel but simple framework named Cascade-CLIP. Our Cascade-CLIP is flexible and can be easily applied to existing zero-shot semantic segmentation methods. Experimental results show that our simple Cascade-CLIP achieves superior zero-shot performance on segmentation benchmarks, like COCO-Stuff, Pascal-VOC, and Pascal-Context. Our code is available at: https://github.com/HVision-NKU/Cascade-CLIP

Jiyao Liu, Jinjie Wei, Wanying Qu et al.

Medical Image Quality Assessment (IQA) serves as the first-mile safety gate for clinical AI, yet existing approaches remain constrained by scalar, score-based metrics and fail to reflect the descriptive, human-like reasoning process central to expert evaluation. To address this gap, we introduce MedQ-Bench, a comprehensive benchmark that establishes a perception-reasoning paradigm for language-based evaluation of medical image quality with Multi-modal Large Language Models (MLLMs). MedQ-Bench defines two complementary tasks: (1) MedQ-Perception, which probes low-level perceptual capability via human-curated questions on fundamental visual attributes; and (2) MedQ-Reasoning, encompassing both no-reference and comparison reasoning tasks, aligning model evaluation with human-like reasoning on image quality. The benchmark spans five imaging modalities and over forty quality attributes, totaling 2,600 perceptual queries and 708 reasoning assessments, covering diverse image sources including authentic clinical acquisitions, images with simulated degradations via physics-based reconstructions, and AI-generated images. To evaluate reasoning ability, we propose a multi-dimensional judging protocol that assesses model outputs along four complementary axes. We further conduct rigorous human-AI alignment validation by comparing LLM-based judgement with radiologists. Our evaluation of 14 state-of-the-art MLLMs demonstrates that models exhibit preliminary but unstable perceptual and reasoning skills, with insufficient accuracy for reliable clinical use. These findings highlight the need for targeted optimization of MLLMs in medical IQA. We hope that MedQ-Bench will catalyze further exploration and unlock the untapped potential of MLLMs for medical image quality evaluation.

Shi-Chen Zhang, Yunheng Li, Yu-Huan Wu et al.

Semantic segmentation is fundamental to vision systems requiring pixel-level scene understanding, yet deploying it on resource-constrained devices demands efficient architectures. Although existing methods achieve real-time inference through lightweight designs, we reveal their inherent limitation: misalignment between class representations and image features caused by a per-pixel classification paradigm. With experimental analysis, we find that this paradigm results in a highly challenging assumption for efficient scenarios: Image pixel features should not vary for the same category in different images. To address this dilemma, we propose a coupled dual-branch offset learning paradigm that explicitly learns feature and class offsets to dynamically refine both class representations and spatial image features. Based on the proposed paradigm, we construct an efficient semantic segmentation network, OffSeg. Notably, the offset learning paradigm can be adopted to existing methods with no additional architectural changes. Extensive experiments on four datasets, including ADE20K, Cityscapes, COCO-Stuff-164K, and Pascal Context, demonstrate consistent improvements with negligible parameters. For instance, on the ADE20K dataset, our proposed offset learning paradigm improves SegFormer-B0, SegNeXt-T, and Mask2Former-Tiny by 2.7%, 1.9%, and 2.6% mIoU, respectively, with only 0.1-0.2M additional parameters required.

Quan-Sheng Zeng, Yunheng Li, Qilong Wang et al.

Visual token compression is critical for Large Vision-Language Models (LVLMs) to efficiently process high-resolution inputs. Existing methods that typically adopt fixed compression ratios cannot adapt to scenes of varying complexity, often causing imprecise pruning that discards informative visual tokens and results in degraded model performance. To address this issue, we introduce a dynamic pruning framework, GlimpsePrune, inspired by human cognition. It takes a data-driven ''glimpse'' and prunes irrelevant visual tokens in a single forward pass before answer generation. This approach prunes 92.6% of visual tokens while on average fully retaining the baseline performance on free-form VQA tasks. The reduced computational cost also enables more effective fine-tuning: an enhanced GlimpsePrune+ achieves 110% of the baseline performance while maintaining a similarly high pruning rate. Our work paves a new way for building more powerful and efficient LVLMs.

Zhong-Yu Li, Yunheng Li, Deng-Ping Fan et al.

Masked image modeling has achieved great success in learning representations but is limited by the huge computational costs. One cost-saving strategy makes the decoder reconstruct only a subset of masked tokens and throw the others, and we refer to this method as partial reconstruction. However, it also degrades the representation quality. Previous methods mitigate this issue by throwing tokens with minimal information using temporal redundancy inaccessible for static images or attention maps that incur extra costs and complexity. To address these limitations, we propose a progressive reconstruction strategy and a furthest sampling strategy to reconstruct those thrown tokens in an extremely lightweight way instead of completely abandoning them. This approach involves all masked tokens in supervision to ensure adequate pre-training, while maintaining the cost-reduction benefits of partial reconstruction. We validate the effectiveness of the proposed method across various existing frameworks. For example, when throwing 50% patches, we can achieve lossless performance of the ViT-B/16 while saving 28% FLOPs and 36% memory usage compared to standard MAE. Our source code will be made publicly available

Yunheng Li, Zhongyu Li, Shanghua Gao et al.

Effectively modeling discriminative spatio-temporal information is essential for segmenting activities in long action sequences. However, we observe that existing methods are limited in weak spatio-temporal modeling capability due to two forms of decoupled modeling: (i) cascaded interaction couples spatial and temporal modeling, which over-smooths motion modeling over the long sequence, and (ii) joint-shared temporal modeling adopts shared weights to model each joint, ignoring the distinct motion patterns of different joints. We propose a Decoupled Spatio-Temporal Framework (DeST) to address the above issues. Firstly, we decouple the cascaded spatio-temporal interaction to avoid stacking multiple spatio-temporal blocks, while achieving sufficient spatio-temporal interaction. Specifically, DeST performs once unified spatial modeling and divides the spatial features into different groups of subfeatures, which then adaptively interact with temporal features from different layers. Since the different sub-features contain distinct spatial semantics, the model could learn the optimal interaction pattern at each layer. Meanwhile, inspired by the fact that different joints move at different speeds, we propose joint-decoupled temporal modeling, which employs independent trainable weights to capture distinctive temporal features of each joint. On four large-scale benchmarks of different scenes, DeST significantly outperforms current state-of-the-art methods with less computational complexity.