Feng Zhao

Kou Shi, Ziao Zhang, Shiting Huang et al.

Large language model (LLM)-based agents have shown strong capabilities in using external tools to solve complex tasks. However, existing evaluations often overlook the temporal dimension of tool use, especially the impact of tool response latency, and are usually limited to single-task settings. In real-world applications, multiple tasks often need to be executed concurrently, and overall efficiency depends on whether an agent can use idle time while waiting for tool responses. We refer to this capability as asynchronous tool calling. To evaluate it, we propose AsyncTool, a benchmark for assessing LLM-based agents in interactive multi-task tool-use environments with delayed tool feedback. AsyncTool presents multiple heterogeneous tasks simultaneously and simulates realistic tool response latency during execution. Using a hybrid data evolution strategy, we construct a diverse asynchronous multitasking dataset that covers multiple scenarios and tool-use patterns. We evaluate models at the step, sub-task, and task levels, and introduce efficiency-oriented metrics to measure task coordination and completion efficiency. Extensive experiments show that delayed tool feedback poses substantial challenges to current agents and leads to clear performance degradation. Models that better coordinate task switching, dependency tracking, and state maintenance achieve stronger performance on AsyncTool. Our analysis identifies key failure modes of current tool-using agents and provides practical insights for designing future systems with stronger temporal reasoning and coordination capabilities.

Naishan Zheng, Man Zhou, Yanmeng Dong et al.

Deep neural networks have achieved remarkable progress in enhancing low-light images by improving their brightness and eliminating noise. However, most existing methods construct end-to-end mapping networks heuristically, neglecting the intrinsic prior of image enhancement task and lacking transparency and interpretability. Although some unfolding solutions have been proposed to relieve these issues, they rely on proximal operator networks that deliver ambiguous and implicit priors. In this work, we propose a paradigm for low-light image enhancement that explores the potential of customized learnable priors to improve the transparency of the deep unfolding paradigm. Motivated by the powerful feature representation capability of Masked Autoencoder (MAE), we customize MAE-based illumination and noise priors and redevelop them from two perspectives: 1) \textbf{structure flow}: we train the MAE from a normal-light image to its illumination properties and then embed it into the proximal operator design of the unfolding architecture; and m2) \textbf{optimization flow}: we train MAE from a normal-light image to its gradient representation and then employ it as a regularization term to constrain noise in the model output. These designs improve the interpretability and representation capability of the model.Extensive experiments on multiple low-light image enhancement datasets demonstrate the superiority of our proposed paradigm over state-of-the-art methods. Code is available at https://github.com/zheng980629/CUE.

Zehui Chen, Zhenyu Li, Shiquan Zhang et al.

Point clouds and RGB images are two general perceptional sources in autonomous driving. The former can provide accurate localization of objects, and the latter is denser and richer in semantic information. Recently, AutoAlign presents a learnable paradigm in combining these two modalities for 3D object detection. However, it suffers from high computational cost introduced by the global-wise attention. To solve the problem, we propose Cross-Domain DeformCAFA module in this work. It attends to sparse learnable sampling points for cross-modal relational modeling, which enhances the tolerance to calibration error and greatly speeds up the feature aggregation across different modalities. To overcome the complex GT-AUG under multi-modal settings, we design a simple yet effective cross-modal augmentation strategy on convex combination of image patches given their depth information. Moreover, by carrying out a novel image-level dropout training scheme, our model is able to infer in a dynamic manner. To this end, we propose AutoAlignV2, a faster and stronger multi-modal 3D detection framework, built on top of AutoAlign. Extensive experiments on nuScenes benchmark demonstrate the effectiveness and efficiency of AutoAlignV2. Notably, our best model reaches 72.4 NDS on nuScenes test leaderboard, achieving new state-of-the-art results among all published multi-modal 3D object detectors. Code will be available at https://github.com/zehuichen123/AutoAlignV2.

Zehui Chen, Zhenyu Li, Shiquan Zhang et al.

3D object detection from multiple image views is a fundamental and challenging task for visual scene understanding. Owing to its low cost and high efficiency, multi-view 3D object detection has demonstrated promising application prospects. However, accurately detecting objects through perspective views is extremely difficult due to the lack of depth information. Current approaches tend to adopt heavy backbones for image encoders, making them inapplicable for real-world deployment. Different from the images, LiDAR points are superior in providing spatial cues, resulting in highly precise localization. In this paper, we explore the incorporation of LiDAR-based detectors for multi-view 3D object detection. Instead of directly training a depth prediction network, we unify the image and LiDAR features in the Bird-Eye-View (BEV) space and adaptively transfer knowledge across non-homogenous representations in a teacher-student paradigm. To this end, we propose \textbf{BEVDistill}, a cross-modal BEV knowledge distillation (KD) framework for multi-view 3D object detection. Extensive experiments demonstrate that the proposed method outperforms current KD approaches on a highly-competitive baseline, BEVFormer, without introducing any extra cost in the inference phase. Notably, our best model achieves 59.4 NDS on the nuScenes test leaderboard, achieving new state-of-the-art in comparison with various image-based detectors. Code will be available at https://github.com/zehuichen123/BEVDistill.

Man Zhou, Hu Yu, Jie Huang et al.

Existing convolutional neural networks widely adopt spatial down-/up-sampling for multi-scale modeling. However, spatial up-sampling operators (\emph{e.g.}, interpolation, transposed convolution, and un-pooling) heavily depend on local pixel attention, incapably exploring the global dependency. In contrast, the Fourier domain obeys the nature of global modeling according to the spectral convolution theorem. Unlike the spatial domain that performs up-sampling with the property of local similarity, up-sampling in the Fourier domain is more challenging as it does not follow such a local property. In this study, we propose a theoretically sound Deep Fourier Up-Sampling (FourierUp) to solve these issues. We revisit the relationships between spatial and Fourier domains and reveal the transform rules on the features of different resolutions in the Fourier domain, which provide key insights for FourierUp's designs. FourierUp as a generic operator consists of three key components: 2D discrete Fourier transform, Fourier dimension increase rules, and 2D inverse Fourier transform, which can be directly integrated with existing networks. Extensive experiments across multiple computer vision tasks, including object detection, image segmentation, image de-raining, image dehazing, and guided image super-resolution, demonstrate the consistent performance gains obtained by introducing our FourierUp.

Lin Chen, Jinsong Li, Xiaoyi Dong et al.

In the realm of large multi-modal models (LMMs), efficient modality alignment is crucial yet often constrained by the scarcity of high-quality image-text data. To address this bottleneck, we introduce the ShareGPT4V dataset, a pioneering large-scale resource featuring 1.2 million highly descriptive captions, which surpasses existing datasets in diversity and information content, covering world knowledge, object properties, spatial relationships, and aesthetic evaluations. Specifically, ShareGPT4V originates from a curated 100K high-quality captions collected from advanced GPT4-Vision and has been expanded to 1.2M with a superb caption model trained on this subset. ShareGPT4V first demonstrates its effectiveness for the Supervised Fine-Tuning (SFT) phase, by substituting an equivalent quantity of detailed captions in existing SFT datasets with a subset of our high-quality captions, significantly enhancing the LMMs like LLaVA-7B, LLaVA-1.5-13B, and Qwen-VL-Chat-7B on the MME and MMBench benchmarks, with respective gains of 222.8/22.0/22.3 and 2.7/1.3/1.5. We further incorporate ShareGPT4V data into both the pre-training and SFT phases, obtaining ShareGPT4V-7B, a superior LMM based on a simple architecture that has remarkable performance across a majority of the multi-modal benchmarks. This project is available at https://ShareGPT4V.github.io to serve as a pivotal resource for advancing the LMMs community.

Hongjing Niu, Hanting Li, Feng Zhao et al.

Domain generalization (DG), aiming to make models work on unseen domains, is a surefire way toward general artificial intelligence. Limited by the scale and diversity of current DG datasets, it is difficult for existing methods to scale to diverse domains in open-world scenarios (e.g., science fiction and pixelate style). Therefore, the source-free domain generalization (SFDG) task is necessary and challenging. To address this issue, we propose an approach based on large-scale vision-language pretraining models (e.g., CLIP), which exploits the extensive domain information embedded in it. The proposed scheme generates diverse prompts from a domain bank that contains many more diverse domains than existing DG datasets. Furthermore, our method yields domain-unified representations from these prompts, thus being able to cope with samples from open-world domains. Extensive experiments on mainstream DG datasets, namely PACS, VLCS, OfficeHome, and DomainNet, show that the proposed method achieves competitive performance compared to state-of-the-art (SOTA) DG methods that require source domain data for training. Besides, we collect a small datasets consists of two domains to evaluate the open-world domain generalization ability of the proposed method. The source code and the dataset will be made publicly available at https://github.com/muse1998/Source-Free-Domain-Generalization

Guangkai Xu, Wei Yin, Hao Chen et al.

3D scene reconstruction is a long-standing vision task. Existing approaches can be categorized into geometry-based and learning-based methods. The former leverages multi-view geometry but can face catastrophic failures due to the reliance on accurate pixel correspondence across views. The latter was proffered to mitigate these issues by learning 2D or 3D representation directly. However, without a large-scale video or 3D training data, it can hardly generalize to diverse real-world scenarios due to the presence of tens of millions or even billions of optimization parameters in the deep network. Recently, robust monocular depth estimation models trained with large-scale datasets have been proven to possess weak 3D geometry prior, but they are insufficient for reconstruction due to the unknown camera parameters, the affine-invariant property, and inter-frame inconsistency. Here, we propose a novel test-time optimization approach that can transfer the robustness of affine-invariant depth models such as LeReS to challenging diverse scenes while ensuring inter-frame consistency, with only dozens of parameters to optimize per video frame. Specifically, our approach involves freezing the pre-trained affine-invariant depth model's depth predictions, rectifying them by optimizing the unknown scale-shift values with a geometric consistency alignment module, and employing the resulting scale-consistent depth maps to robustly obtain camera poses and achieve dense scene reconstruction, even in low-texture regions. Experiments show that our method achieves state-of-the-art cross-dataset reconstruction on five zero-shot testing datasets.

Naishan Zheng, Jie Huang, Qi Zhu et al.

Low-light image enhancement is an inherently subjective process whose targets vary with the user's aesthetic. Motivated by this, several personalized enhancement methods have been investigated. However, the enhancement process based on user preferences in these techniques is invisible, i.e., a "black box". In this work, we propose an intelligible unsupervised personalized enhancer (iUPEnhancer) for low-light images, which establishes the correlations between the low-light and the unpaired reference images with regard to three user-friendly attributions (brightness, chromaticity, and noise). The proposed iUP-Enhancer is trained with the guidance of these correlations and the corresponding unsupervised loss functions. Rather than a "black box" process, our iUP-Enhancer presents an intelligible enhancement process with the above attributions. Extensive experiments demonstrate that the proposed algorithm produces competitive qualitative and quantitative results while maintaining excellent flexibility and scalability. This can be validated by personalization with single/multiple references, cross-attribution references, or merely adjusting parameters.

Keyu Yan, Man Zhou, Jie Huang et al.

Panchromatic (PAN) and multi-spectral (MS) image fusion, named Pan-sharpening, refers to super-resolve the low-resolution (LR) multi-spectral (MS) images in the spatial domain to generate the expected high-resolution (HR) MS images, conditioning on the corresponding high-resolution PAN images. In this paper, we present a simple yet effective \textit{alternating reverse filtering network} for pan-sharpening. Inspired by the classical reverse filtering that reverses images to the status before filtering, we formulate pan-sharpening as an alternately iterative reverse filtering process, which fuses LR MS and HR MS in an interpretable manner. Different from existing model-driven methods that require well-designed priors and degradation assumptions, the reverse filtering process avoids the dependency on pre-defined exact priors. To guarantee the stability and convergence of the iterative process via contraction mapping on a metric space, we develop the learnable multi-scale Gaussian kernel module, instead of using specific filters. We demonstrate the theoretical feasibility of such formulations. Extensive experiments on diverse scenes to thoroughly verify the performance of our method, significantly outperforming the state of the arts.

Hu Yu, Li Shen, Jie Huang et al.

Denoising diffusion models have emerged as a dominant approach for image generation, however they still suffer from slow convergence in training and color shift issues in sampling. In this paper, we identify that these obstacles can be largely attributed to bias and suboptimality inherent in the default training paradigm of diffusion models. Specifically, we offer theoretical insights that the prevailing constant loss weight strategy in $ε$-prediction of diffusion models leads to biased estimation during the training phase, hindering accurate estimations of original images. To address the issue, we propose a simple but effective weighting strategy derived from the unlocked biased part. Furthermore, we conduct a comprehensive and systematic exploration, unraveling the inherent bias problem in terms of its existence, impact and underlying reasons. These analyses contribute to advancing the understanding of diffusion models. Empirical results demonstrate that our method remarkably elevates sample quality and displays improved efficiency in both training and sampling processes, by only adjusting loss weighting strategy. The code is released publicly at \url{https://github.com/yuhuUSTC/Debias}

Hu Yu, Jie Huang, Kaiwen Zheng et al.

Image dehazing is quite challenging in dense-haze scenarios, where quite less original information remains in the hazy image. Though previous methods have made marvelous progress, they still suffer from information loss in content and color in dense-haze scenarios. The recently emerged Denoising Diffusion Probabilistic Model (DDPM) exhibits strong generation ability, showing potential for solving this problem. However, DDPM fails to consider the physics property of dehazing task, limiting its information completion capacity. In this work, we propose DehazeDDPM: A DDPM-based and physics-aware image dehazing framework that applies to complex hazy scenarios. Specifically, DehazeDDPM works in two stages. The former stage physically models the dehazing task with the Atmospheric Scattering Model (ASM), pulling the distribution closer to the clear data and endowing DehazeDDPM with fog-aware ability. The latter stage exploits the strong generation ability of DDPM to compensate for the haze-induced huge information loss, by working in conjunction with the physical modelling. Extensive experiments demonstrate that our method attains state-of-the-art performance on both synthetic and real-world hazy datasets.

Hu Yu, Jie Huang, Yajing Liu et al.

Deep learning-based source dehazing methods trained on synthetic datasets have achieved remarkable performance but suffer from dramatic performance degradation on real hazy images due to domain shift. Although certain Domain Adaptation (DA) dehazing methods have been presented, they inevitably require access to the source dataset to reduce the gap between the source synthetic and target real domains. To address these issues, we present a novel Source-Free Unsupervised Domain Adaptation (SFUDA) image dehazing paradigm, in which only a well-trained source model and an unlabeled target real hazy dataset are available. Specifically, we devise the Domain Representation Normalization (DRN) module to make the representation of real hazy domain features match that of the synthetic domain to bridge the gaps. With our plug-and-play DRN module, unlabeled real hazy images can adapt existing well-trained source networks. Besides, the unsupervised losses are applied to guide the learning of the DRN module, which consists of frequency losses and physical prior losses. Frequency losses provide structure and style constraints, while the prior loss explores the inherent statistic property of haze-free images. Equipped with our DRN module and unsupervised loss, existing source dehazing models are able to dehaze unlabeled real hazy images. Extensive experiments on multiple baselines demonstrate the validity and superiority of our method visually and quantitatively.

Wenxuan Huang, Yu Zeng, Qiuchen Wang et al.

Multimodal large language models (MLLMs) have achieved remarkable success across a broad range of vision tasks. However, constrained by the capacity of their internal world knowledge, prior work has proposed augmenting MLLMs by ``reasoning-then-tool-call'' for visual and textual search engines to obtain substantial gains on tasks requiring extensive factual information. However, these approaches typically define multimodal search in a naive setting, assuming that a single full-level or entity-level image query and few text query suffices to retrieve the key evidence needed to answer the question, which is unrealistic in real-world scenarios with substantial visual noise. Moreover, they are often limited in the reasoning depth and search breadth, making it difficult to solve complex questions that require aggregating evidence from diverse visual and textual sources. Building on this, we propose Vision-DeepResearch, which proposes one new multimodal deep-research paradigm, i.e., performs multi-turn, multi-entity and multi-scale visual and textual search to robustly hit real-world search engines under heavy noise. Our Vision-DeepResearch supports dozens of reasoning steps and hundreds of engine interactions, while internalizing deep-research capabilities into the MLLM via cold-start supervision and RL training, resulting in a strong end-to-end multimodal deep-research MLLM. It substantially outperforming existing multimodal deep-research MLLMs, and workflows built on strong closed-source foundation model such as GPT-5, Gemini-2.5-pro and Claude-4-Sonnet. The code will be released in https://github.com/Osilly/Vision-DeepResearch.

Hanting Li, Hongjing Niu, Zhaoqing Zhu et al.

Compared with the image-based static facial expression recognition (SFER) task, the dynamic facial expression recognition (DFER) task based on video sequences is closer to the natural expression recognition scene. However, DFER is often more challenging. One of the main reasons is that video sequences often contain frames with different expression intensities, especially for the facial expressions in the real-world scenarios, while the images in SFER frequently present uniform and high expression intensities. However, if the expressions with different intensities are treated equally, the features learned by the networks will have large intra-class and small inter-class differences, which is harmful to DFER. To tackle this problem, we propose the global convolution-attention block (GCA) to rescale the channels of the feature maps. In addition, we introduce the intensity-aware loss (IAL) in the training process to help the network distinguish the samples with relatively low expression intensities. Experiments on two in-the-wild dynamic facial expression datasets (i.e., DFEW and FERV39k) indicate that our method outperforms the state-of-the-art DFER approaches. The source code will be made publicly available.

Qianhao Yu, Naishan Zheng, Jie Huang et al.

The key to shadow removal is recovering the contents of the shadow regions with the guidance of the non-shadow regions. Due to the inadequate long-range modeling, the CNN-based approaches cannot thoroughly investigate the information from the non-shadow regions. To solve this problem, we propose a novel cleanness-navigated-shadow network (CNSNet), with a shadow-oriented adaptive normalization (SOAN) module and a shadow-aware aggregation with transformer (SAAT) module based on the shadow mask. Under the guidance of the shadow mask, the SOAN module formulates the statistics from the non-shadow region and adaptively applies them to the shadow region for region-wise restoration. The SAAT module utilizes the shadow mask to precisely guide the restoration of each shadowed pixel by considering the highly relevant pixels from the shadow-free regions for global pixel-wise restoration. Extensive experiments on three benchmark datasets (ISTD, ISTD+, and SRD) show that our method achieves superior de-shadowing performance.

Zehui Chen, Zhenyu Li, Shiquan Zhang et al.

3D object detection from multiple image views is a fundamental and challenging task for visual scene understanding. Due to its low cost and high efficiency, multi-view 3D object detection has demonstrated promising application prospects. However, accurately detecting objects through perspective views in the 3D space is extremely difficult due to the lack of depth information. Recently, DETR3D introduces a novel 3D-2D query paradigm in aggregating multi-view images for 3D object detection and achieves state-of-the-art performance. In this paper, with intensive pilot experiments, we quantify the objects located at different regions and find that the "truncated instances" (i.e., at the border regions of each image) are the main bottleneck hindering the performance of DETR3D. Although it merges multiple features from two adjacent views in the overlapping regions, DETR3D still suffers from insufficient feature aggregation, thus missing the chance to fully boost the detection performance. In an effort to tackle the problem, we propose Graph-DETR3D to automatically aggregate multi-view imagery information through graph structure learning (GSL). It constructs a dynamic 3D graph between each object query and 2D feature maps to enhance the object representations, especially at the border regions. Besides, Graph-DETR3D benefits from a novel depth-invariant multi-scale training strategy, which maintains the visual depth consistency by simultaneously scaling the image size and the object depth. Extensive experiments on the nuScenes dataset demonstrate the effectiveness and efficiency of our Graph-DETR3D. Notably, our best model achieves 49.5 NDS on the nuScenes test leaderboard, achieving new state-of-the-art in comparison with various published image-view 3D object detectors.

Qingnan Ren, Shun Zou, Shiting Huang et al.

As autonomous coding agents become capable of handling increasingly long-horizon tasks, they have gradually demonstrated the potential to complete end-to-end software development. Although existing benchmarks have recently evolved from localized code editing to from-scratch project generation, they remain confined to structurally simplified, single-stack applications. Consequently, they fail to capture the heterogeneous environments, full-stack orchestration, and system-level complexity of real enterprise Software as a Service (SaaS) systems, leaving a critical gap in assessing agents under realistic engineering constraints. To fill this gap, we introduce SaaSBench, the first benchmark designed to explore the boundaries of AI agents in enterprise SaaS engineering. Spanning 30 complex tasks across 6 SaaS domains with 5,370 validation nodes, it incorporates 8 programming languages, 6 databases, and 13 frameworks to meticulously mirror real-world software heterogeneity. Furthermore, we design a dependency-aware hybrid evaluation paradigm tailored for complex systems with long horizons and multi-component coupling, enabling fine-grained, reproducible assessment. Crucially, our extensive experiments reveal a striking insight: the primary bottleneck for state-of-the-art agents is not generating isolated code logic, but successfully configuring and integrating a multi-component system. Over 95\% of task failures occur before agents even reach deep business logic, with models often falling victim to overconfidence and prematurely halting during foundational system setup, or getting trapped in ineffective debugging loops. We hope SaaSBench serves as a practical and challenging testbed to drive the evolution of reliable, system-level coding agents. The code is available at \url{https://github.com/ShadeCloak/SaaSbench}.

Hanting Li, Hongjing Niu, Zhaoqing Zhu et al.

Facial expression recognition (FER) is an essential task for understanding human behaviors. As one of the most informative behaviors of humans, facial expressions are often compound and variable, which is manifested by the fact that different people may express the same expression in very different ways. However, most FER methods still use one-hot or soft labels as the supervision, which lack sufficient semantic descriptions of facial expressions and are less interpretable. Recently, contrastive vision-language pre-training (VLP) models (e.g., CLIP) use text as supervision and have injected new vitality into various computer vision tasks, benefiting from the rich semantics in text. Therefore, in this work, we propose CLIPER, a unified framework for both static and dynamic facial Expression Recognition based on CLIP. Besides, we introduce multiple expression text descriptors (METD) to learn fine-grained expression representations that make CLIPER more interpretable. We conduct extensive experiments on several popular FER benchmarks and achieve state-of-the-art performance, which demonstrates the effectiveness of CLIPER.

Jiahao Chang, Shuo Wang, Haiming Xu et al.

Transformer-based detectors (DETRs) are becoming popular for their simple framework, but the large model size and heavy time consumption hinder their deployment in the real world. While knowledge distillation (KD) can be an appealing technique to compress giant detectors into small ones for comparable detection performance and low inference cost. Since DETRs formulate object detection as a set prediction problem, existing KD methods designed for classic convolution-based detectors may not be directly applicable. In this paper, we propose DETRDistill, a novel knowledge distillation method dedicated to DETR-families. Specifically, we first design a Hungarian-matching logits distillation to encourage the student model to have the exact predictions as that of teacher DETRs. Next, we propose a target-aware feature distillation to help the student model learn from the object-centric features of the teacher model. Finally, in order to improve the convergence rate of the student DETR, we introduce a query-prior assignment distillation to speed up the student model learning from well-trained queries and stable assignment of the teacher model. Extensive experimental results on the COCO dataset validate the effectiveness of our approach. Notably, DETRDistill consistently improves various DETRs by more than 2.0 mAP, even surpassing their teacher models.

Bowen Zhang, Yiji Cheng, Chunyu Wang et al.

We present RodinHD, which can generate high-fidelity 3D avatars from a portrait image. Existing methods fail to capture intricate details such as hairstyles which we tackle in this paper. We first identify an overlooked problem of catastrophic forgetting that arises when fitting triplanes sequentially on many avatars, caused by the MLP decoder sharing scheme. To overcome this issue, we raise a novel data scheduling strategy and a weight consolidation regularization term, which improves the decoder's capability of rendering sharper details. Additionally, we optimize the guiding effect of the portrait image by computing a finer-grained hierarchical representation that captures rich 2D texture cues, and injecting them to the 3D diffusion model at multiple layers via cross-attention. When trained on 46K avatars with a noise schedule optimized for triplanes, the resulting model can generate 3D avatars with notably better details than previous methods and can generalize to in-the-wild portrait input.

Hanting Li, Mingzhe Sui, Zhaoqing Zhu et al.

Dynamic facial expression recognition (DFER) in the wild is an extremely challenging task, due to a large number of noisy frames in the video sequences. Previous works focus on extracting more discriminative features, but ignore distinguishing the key frames from the noisy frames. To tackle this problem, we propose a noise-robust dynamic facial expression recognition network (NR-DFERNet), which can effectively reduce the interference of noisy frames on the DFER task. Specifically, at the spatial stage, we devise a dynamic-static fusion module (DSF) that introduces dynamic features to static features for learning more discriminative spatial features. To suppress the impact of target irrelevant frames, we introduce a novel dynamic class token (DCT) for the transformer at the temporal stage. Moreover, we design a snippet-based filter (SF) at the decision stage to reduce the effect of too many neutral frames on non-neutral sequence classification. Extensive experimental results demonstrate that our NR-DFERNet outperforms the state-of-the-art methods on both the DFEW and AFEW benchmarks.

Shuo Wang, Xinhai Zhao, Hai-Ming Xu et al.

Multi-view 3D object detection (MV3D-Det) in Bird-Eye-View (BEV) has drawn extensive attention due to its low cost and high efficiency. Although new algorithms for camera-only 3D object detection have been continuously proposed, most of them may risk drastic performance degradation when the domain of input images differs from that of training. In this paper, we first analyze the causes of the domain gap for the MV3D-Det task. Based on the covariate shift assumption, we find that the gap mainly attributes to the feature distribution of BEV, which is determined by the quality of both depth estimation and 2D image's feature representation. To acquire a robust depth prediction, we propose to decouple the depth estimation from the intrinsic parameters of the camera (i.e. the focal length) through converting the prediction of metric depth to that of scale-invariant depth and perform dynamic perspective augmentation to increase the diversity of the extrinsic parameters (i.e. the camera poses) by utilizing homography. Moreover, we modify the focal length values to create multiple pseudo-domains and construct an adversarial training loss to encourage the feature representation to be more domain-agnostic. Without bells and whistles, our approach, namely DG-BEV, successfully alleviates the performance drop on the unseen target domain without impairing the accuracy of the source domain. Extensive experiments on various public datasets, including Waymo, nuScenes, and Lyft, demonstrate the generalization and effectiveness of our approach. To the best of our knowledge, this is the first systematic study to explore a domain generalization method for MV3D-Det.

Yuanyang Yin, Yaqi Zhao, Yajie Zhang et al.

Multimodal Large Language Models (MLLMs) have demonstrated remarkable capabilities by integrating visual and textual inputs, yet modality alignment remains one of the most challenging aspects. Current MLLMs typically rely on simple adapter architectures and pretraining approaches to bridge vision encoders with large language models (LLM), guided by image-level supervision. We identify this paradigm often leads to suboptimal alignment between modalities, significantly constraining the LLM's ability to properly interpret and reason with visual features particularly for smaller language models. This limitation degrades overall performance-particularly for smaller language models where capacity constraints are more pronounced and adaptation capabilities are limited. To address this fundamental limitation, we propose Supervised Embedding Alignment (SEA), a token-level supervision alignment method that enables more precise visual-text alignment during pretraining. SEA introduces minimal computational overhead while preserving language capabilities and substantially improving cross-modal understanding. Our comprehensive analyses reveal critical insights into the adapter's role in multimodal integration, and extensive experiments demonstrate that SEA consistently improves performance across various model sizes, with smaller models benefiting the most (average performance gain of 7.61% for Gemma-2B). This work establishes a foundation for developing more effective alignment strategies for future multimodal systems.

Zehui Chen, Kuikun Liu, Qiuchen Wang et al.

Information seeking and integration is a complex cognitive task that consumes enormous time and effort. Inspired by the remarkable progress of Large Language Models, recent works attempt to solve this task by combining LLMs and search engines. However, these methods still obtain unsatisfying performance due to three challenges: (1) complex requests often cannot be accurately and completely retrieved by the search engine once (2) corresponding information to be integrated is spread over multiple web pages along with massive noise, and (3) a large number of web pages with long contents may quickly exceed the maximum context length of LLMs. Inspired by the cognitive process when humans solve these problems, we introduce MindSearch to mimic the human minds in web information seeking and integration, which can be instantiated by a simple yet effective LLM-based multi-agent framework. The WebPlanner models the human mind of multi-step information seeking as a dynamic graph construction process: it decomposes the user query into atomic sub-questions as nodes in the graph and progressively extends the graph based on the search result from WebSearcher. Tasked with each sub-question, WebSearcher performs hierarchical information retrieval with search engines and collects valuable information for WebPlanner. The multi-agent design of MindSearch enables the whole framework to seek and integrate information parallelly from larger-scale (e.g., more than 300) web pages in 3 minutes, which is worth 3 hours of human effort. MindSearch demonstrates significant improvement in the response quality in terms of depth and breadth, on both close-set and open-set QA problems. Besides, responses from MindSearch based on InternLM2.5-7B are preferable by humans to ChatGPT-Web and Perplexity.ai applications, which implies that MindSearch can already deliver a competitive solution to the proprietary AI search engine.

Jinghao Zhang, Feng Zhao

Learning to restore multiple image degradations within a single model is quite beneficial for real-world applications. Nevertheless, existing works typically concentrate on regarding each degradation independently, while their relationship has been less exploited to ensure the synergistic learning. To this end, we revisit the diverse degradations through the lens of singular value decomposition, with the observation that the decomposed singular vectors and singular values naturally undertake the different types of degradation information, dividing various restoration tasks into two groups, \ie, singular vector dominated and singular value dominated. The above analysis renders a more unified perspective to ascribe the diverse degradations, compared to previous task-level independent learning. The dedicated optimization of degraded singular vectors and singular values inherently utilizes the potential relationship among diverse restoration tasks, attributing to the Decomposition Ascribed Synergistic Learning (DASL). Specifically, DASL comprises two effective operators, namely, Singular VEctor Operator (SVEO) and Singular VAlue Operator (SVAO), to favor the decomposed optimization, which can be lightly integrated into existing image restoration backbone. Moreover, the congruous decomposition loss has been devised for auxiliary. Extensive experiments on blended five image restoration tasks demonstrate the effectiveness of our method.

Jinghao Zhang, Ruilin Li, Feng Zhao et al.

Reinforcement learning with verifiable rewards (RLVR) has become a core technique for post-training of Large Language Models (LLMs). While policy optimization is driven by all sampled tokens under a globally broadcast scalar reward, the heterogeneous policy behaviors exhibited along trajectories are largely overlooked without differentiation. Existing works address this by credit allocation, including token-level advantage reweighting, and selective token optimization, however, the allocation criterion are principally stagnant throughout training, limiting resilient policy evolution. In this work, we argue that \textit{when} learning signals are scheduled can be as important as \textit{where} they are allocated across tokens, and introduce the temporal dimension that scheduling the credit allocation criteria over the course of RLVR optimization. We find that prioritizing targeted tokens emphasized with specific policy behaviors, and gradually attenuating toward general optimization leads to more stable and efficient learning dynamics. Furthermore, we show that simple trajectory percentiles provide a natural perspective for distinguishing policy behaviors, and works effectively with temporal scheduling. Our analysis reveals that standard optimization substantially sacrifices policy entropy when simultaneously accommodating heterogeneous behaviors, whereas temporal scheduling yields healthier policy evolution dynamics. Experiments across mathematical and general reasoning benchmarks demonstrate consistent improvements, suggesting that temporal scheduling constitutes a promising optimization dimension.

Thanveer Shaik, Xiaohui Tao, Lin Li et al.

Effective patient monitoring is vital for timely interventions and improved healthcare outcomes. Traditional monitoring systems often struggle to handle complex, dynamic environments with fluctuating vital signs, leading to delays in identifying critical conditions. To address this challenge, we propose a novel AI-driven patient monitoring framework using multi-agent deep reinforcement learning (DRL). Our approach deploys multiple learning agents, each dedicated to monitoring a specific physiological feature, such as heart rate, respiration, and temperature. These agents interact with a generic healthcare monitoring environment, learn the patients' behavior patterns, and make informed decisions to alert the corresponding Medical Emergency Teams (METs) based on the level of emergency estimated. In this study, we evaluate the performance of the proposed multi-agent DRL framework using real-world physiological and motion data from two datasets: PPG-DaLiA and WESAD. We compare the results with several baseline models, including Q-Learning, PPO, Actor-Critic, Double DQN, and DDPG, as well as monitoring frameworks like WISEML and CA-MAQL. Our experiments demonstrate that the proposed DRL approach outperforms all other baseline models, achieving more accurate monitoring of patient's vital signs. Furthermore, we conduct hyperparameter optimization to fine-tune the learning process of each agent. By optimizing hyperparameters, we enhance the learning rate and discount factor, thereby improving the agents' overall performance in monitoring patient health status.

Zhen Zou, Xiaoxiao Ma, Mingde Yao et al.

Autoregressive (AR)-Diffusion hybrid paradigms combine AR's structured semantic modeling with diffusion's high-fidelity synthesis, yet suffer from a dual speed bottleneck: the sequential AR stage and the iterative multi-step denoising of the diffusion vision decode stage. Existing methods address each in isolation without a unified principle design. We observe that the per-position \emph{prediction entropy} of continuous-space AR models naturally encodes spatially varying generation uncertainty, which simultaneously governing draft prediction quality in the AR stage and reflecting the corrective effort required by vision decoding stage, which is not fully explored before. Since entropy is inherently tied to both bottlenecks, it serves as a natural unifying signal for joint acceleration. In this work, we propose \textbf{Drift-AR}, which leverages entropy signal to accelerate both stages: 1) for AR acceleration, we introduce Entropy-Informed Speculative Decoding that align draft--target entropy distributions via a causal-normalized entropy loss, resolving the entropy mismatch that causes excessive draft rejection; 2) for visual decoder acceleration, we reinterpret entropy as the \emph{physical variance} of the initial state for an anti-symmetric drifting field -- high-entropy positions activate stronger drift toward the data manifold while low-entropy positions yield vanishing drift -- enabling single-step (1-NFE) decoding without iterative denoising or distillation. Moreover, both stages share the same entropy signal, which is computed once with no extra cost. Experiments on MAR, TransDiff, and NextStep-1 demonstrate 3.8--5.5$\times$ speedup with genuine 1-NFE decoding, matching or surpassing original quality. Code will be available at https://github.com/aSleepyTree/Drift-AR.

Deyi Ji, Feng Zhao, Hongtao Lu

Most existing ultra-high resolution (UHR) segmentation methods always struggle in the dilemma of balancing memory cost and local characterization accuracy, which are both taken into account in our proposed Guided Patch-Grouping Wavelet Transformer (GPWFormer) that achieves impressive performances. In this work, GPWFormer is a Transformer ($\mathcal{T}$)-CNN ($\mathcal{C}$) mutual leaning framework, where $\mathcal{T}$ takes the whole UHR image as input and harvests both local details and fine-grained long-range contextual dependencies, while $\mathcal{C}$ takes downsampled image as input for learning the category-wise deep context. For the sake of high inference speed and low computation complexity, $\mathcal{T}$ partitions the original UHR image into patches and groups them dynamically, then learns the low-level local details with the lightweight multi-head Wavelet Transformer (WFormer) network. Meanwhile, the fine-grained long-range contextual dependencies are also captured during this process, since patches that are far away in the spatial domain can also be assigned to the same group. In addition, masks produced by $\mathcal{C}$ are utilized to guide the patch grouping process, providing a heuristics decision. Moreover, the congruence constraints between the two branches are also exploited to maintain the spatial consistency among the patches. Overall, we stack the multi-stage process in a pyramid way. Experiments show that GPWFormer outperforms the existing methods with significant improvements on five benchmark datasets.

Feng Zhao, Ziqi Zhang, Donglin Wang

The knowledge graph (KG) is an essential form of knowledge representation that has grown in prominence in recent years. Because it concentrates on nominal entities and their relationships, traditional knowledge graphs are static and encyclopedic in nature. On this basis, event knowledge graph (Event KG) models the temporal and spatial dynamics by text processing to facilitate downstream applications, such as question-answering, recommendation and intelligent search. Existing KG research, on the other hand, mostly focuses on text processing and static facts, ignoring the vast quantity of dynamic behavioral information included in photos, movies, and pre-trained neural networks. In addition, no effort has been done to include behavioral intelligence information into the knowledge graph for deep reinforcement learning (DRL) and robot learning. In this paper, we propose a novel dynamic knowledge and skill graph (KSG), and then we develop a basic and specific KSG based on CN-DBpedia. The nodes are divided into entity and attribute nodes, with entity nodes containing the agent, environment, and skill (DRL policy or policy representation), and attribute nodes containing the entity description, pre-train network, and offline dataset. KSG can search for different agents' skills in various environments and provide transferable information for acquiring new skills. This is the first study that we are aware of that looks into dynamic KSG for skill retrieval and learning. Extensive experimental results on new skill learning show that KSG boosts new skill learning efficiency.

Zhen Zou, Xiaoxiao Ma, Jie Huang et al.

Autoregressive(AR)-diffusion hybrid paradigms combine AR's structured modeling with diffusion's photorealistic synthesis, yet suffer from high latency due to sequential AR generation and iterative denoising. In this work, we tackle this bottleneck and propose a unified AR-diffusion framework Fast-ARDiff that jointly optimizes both components, accelerating AR speculative decoding while simultaneously facilitating faster diffusion decoding. Specifically: (1) The entropy-informed speculative strategy encourages draft model to produce higher-entropy representations aligned with target model's entropy characteristics, mitigating entropy mismatch and high rejection rates caused by draft overconfidence. (2) For diffusion decoding, rather than treating it as an independent module, we integrate it into the same end-to-end framework using a dynamic scheduler that prioritizes AR optimization to guide the diffusion part in further steps. The diffusion part is optimized through a joint distillation framework combining trajectory and distribution matching, ensuring stable training and high-quality synthesis with extremely few steps. During inference, shallow feature entropy from AR module is used to pre-filter low-entropy drafts, avoiding redundant computation and improving latency. Fast-ARDiff achieves state-of-the-art acceleration across diverse models: on ImageNet 256$\times$256, TransDiff attains 4.3$\times$ lossless speedup, and NextStep-1 achieves 3$\times$ acceleration on text-conditioned generation. Code will be available at https://github.com/aSleepyTree/Fast-ARDiff.

Mingzhe Sui, Hanting Li, Zhaoqing Zhu et al.

2D+3D facial expression recognition (FER) can effectively cope with illumination changes and pose variations by simultaneously merging 2D texture and more robust 3D depth information. Most deep learning-based approaches employ the simple fusion strategy that concatenates the multimodal features directly after fully-connected layers, without considering the different degrees of significance for each modality. Meanwhile, how to focus on both 2D and 3D local features in salient regions is still a great challenge. In this letter, we propose the adaptive fusion network with masks (AFNet-M) for 2D+3D FER. To enhance 2D and 3D local features, we take the masks annotating salient regions of the face as prior knowledge and design the mask attention module (MA) which can automatically learn two modulation vectors to adjust the feature maps. Moreover, we introduce a novel fusion strategy that can perform adaptive fusion at convolutional layers through the designed importance weights computing module (IWC). Experimental results demonstrate that our AFNet-M achieves the state-of-the-art performance on BU-3DFE and Bosphorus datasets and requires fewer parameters in comparison with other models.

Zaibin Zhang, Yongting Zhang, Lijun Li et al.

Multi-agent systems, when enhanced with Large Language Models (LLMs), exhibit profound capabilities in collective intelligence. However, the potential misuse of this intelligence for malicious purposes presents significant risks. To date, comprehensive research on the safety issues associated with multi-agent systems remains limited. In this paper, we explore these concerns through the innovative lens of agent psychology, revealing that the dark psychological states of agents constitute a significant threat to safety. To tackle these concerns, we propose a comprehensive framework (PsySafe) grounded in agent psychology, focusing on three key areas: firstly, identifying how dark personality traits in agents can lead to risky behaviors; secondly, evaluating the safety of multi-agent systems from the psychological and behavioral perspectives, and thirdly, devising effective strategies to mitigate these risks. Our experiments reveal several intriguing phenomena, such as the collective dangerous behaviors among agents, agents' self-reflection when engaging in dangerous behavior, and the correlation between agents' psychological assessments and dangerous behaviors. We anticipate that our framework and observations will provide valuable insights for further research into the safety of multi-agent systems. We will make our data and code publicly accessible at https://github.com/AI4Good24/PsySafe.

Zehui Chen, Weihua Du, Wenwei Zhang et al. · cmu

Large language models (LLM) have achieved remarkable performance on various NLP tasks and are augmented by tools for broader applications. Yet, how to evaluate and analyze the tool-utilization capability of LLMs is still under-explored. In contrast to previous works that evaluate models holistically, we comprehensively decompose the tool utilization into multiple sub-processes, including instruction following, planning, reasoning, retrieval, understanding, and review. Based on that, we further introduce T-Eval to evaluate the tool utilization capability step by step. T-Eval disentangles the tool utilization evaluation into several sub-domains along model capabilities, facilitating the inner understanding of both holistic and isolated competency of LLMs. We conduct extensive experiments on T-Eval and in-depth analysis of various LLMs. T-Eval not only exhibits consistency with the outcome-oriented evaluation but also provides a more fine-grained analysis of the capabilities of LLMs, providing a new perspective in LLM evaluation on tool-utilization ability. The benchmark will be available at https://github.com/open-compass/T-Eval.

Qisheng Su, Zhen Fang, Shiting Huang et al.

Recent development of agents has renewed demand for long-context reasoning capacity of LLMs. However, training LLMs for this capacity requires costly long-document curation or heuristic context synthesis. We observe that agents produce massive trajectories when solving problems, invoking tools and receiving environment observations across many turns. The evidence needed to answer the original question is thus scattered throughout these turns, requiring integration of distant context segments. Nevertheless, standard agent SFT masks tool responses and only trains turn-level tool selection, creating a supervision blind spot where these scattered signals go unused. We propose Agent Context Compilation (ACC), which converts trajectories from search, software engineering, and database querying agents into long-context QA pairs that combine the original question with tool responses and environment observations gathered across multiple turns, training the model to answer directly without tool use. This makes the dependencies between the question and the evidence explicit, enabling direct supervision of long-context reasoning over distant segments without additional annotation. ACC is a simple but effective approach that can be combined with any existing long-context extension or training method, providing scalable supervised fine-tuning data. We validate ACC on long-range dependency modeling tasks through MRCR and GraphWalks, challenging benchmarks requiring cross-turn coreference resolution and graph traversal over extended contexts. Training Qwen3-30B-A3B with ACC achieves 68.3 on MRCR (+18.1) and 77.5 on GraphWalks (+7.6), results comparable to Qwen3-235B-A22B, while preserving general capabilities on GPQA, MMLU-Pro, AIME, and IFEval. Further mechanism analysis reveals that the ACC-trained model exhibits task-adaptive attention restructuring and expert specialization.

Ziao Zhang, Kou Shi, Shiting Huang et al.

As the capability frontier of autonomous agents continues to expand, they are increasingly able to complete specialized tasks through plug-and-play external skills. Yet current benchmarks mostly test whether models can use provided skills, leaving open whether they can discover skills from experience, repair them after failure, and maintain a coherent library over time. We introduce SkillFlow, a benchmark of 166 tasks across 20 families in which task construction within each family follows a Domain-Agnostic Execution Flow (DAEF) that defines an agent workflow framework, allowing these tasks to share a consistent workflow. Agents are evaluated under an Agentic Lifelong Learning protocol in which they begin without skills, solve tasks sequentially within each family, externalize lessons through trajectory- and rubric-driven skill patches, and carry the updated library forward. Experiments reveal a substantial capability gap. For Claude Opus 4.6, lifelong skill evolution improves task success from 62.65% to 71.08% (+8.43 points). However, high skill usage does not necessarily imply high utility: Kimi K2.5 gains only +0.60 points despite 66.87% skill usage, while Qwen-Coder-Next reaches only a 44.58% task completion rate and still regresses relative to the vanilla setting. SkillFlow contributes a structured testbed for this direction and an in-depth empirical analysis of skill discovery, patching, transfer, and their failure modes under lifelong evaluation.

Qiuchen Wang, Ruixue Ding, Zehui Chen et al.

Understanding information from visually rich documents remains a significant challenge for traditional Retrieval-Augmented Generation (RAG) methods. Existing benchmarks predominantly focus on image-based question answering (QA), overlooking the fundamental challenges of efficient retrieval, comprehension, and reasoning within dense visual documents. To bridge this gap, we introduce ViDoSeek, a novel dataset designed to evaluate RAG performance on visually rich documents requiring complex reasoning. Based on it, we identify key limitations in current RAG approaches: (i) purely visual retrieval methods struggle to effectively integrate both textual and visual features, and (ii) previous approaches often allocate insufficient reasoning tokens, limiting their effectiveness. To address these challenges, we propose ViDoRAG, a novel multi-agent RAG framework tailored for complex reasoning across visual documents. ViDoRAG employs a Gaussian Mixture Model (GMM)-based hybrid strategy to effectively handle multi-modal retrieval. To further elicit the model's reasoning capabilities, we introduce an iterative agent workflow incorporating exploration, summarization, and reflection, providing a framework for investigating test-time scaling in RAG domains. Extensive experiments on ViDoSeek validate the effectiveness and generalization of our approach. Notably, ViDoRAG outperforms existing methods by over 10% on the competitive ViDoSeek benchmark. The code is available at https://github.com/Alibaba-NLP/ViDoRAG.

Qiuchen Wang, Ruixue Ding, Yu Zeng et al.

Effectively retrieving, reasoning and understanding visually rich information remains a challenge for RAG methods. Traditional text-based methods cannot handle visual-related information. On the other hand, current vision-based RAG approaches are often limited by fixed pipelines and frequently struggle to reason effectively due to the insufficient activation of the fundamental capabilities of models. As RL has been proven to be beneficial for model reasoning, we introduce VRAG-RL, a novel RL framework tailored for complex reasoning across visually rich information. With this framework, VLMs interact with search engines, autonomously sampling single-turn or multi-turn reasoning trajectories with the help of visual perception tokens and undergoing continual optimization based on these samples. Our approach highlights key limitations of RL in RAG domains: (i) Prior Multi-modal RAG approaches tend to merely incorporate images into the context, leading to insufficient reasoning token allocation and neglecting visual-specific perception; and (ii) When models interact with search engines, their queries often fail to retrieve relevant information due to the inability to articulate requirements, thereby leading to suboptimal performance. To address these challenges, we define an action space tailored for visually rich inputs, with actions including cropping and scaling, allowing the model to gather information from a coarse-to-fine perspective. Furthermore, to bridge the gap between users' original inquiries and the retriever, we employ a simple yet effective reward that integrates query rewriting and retrieval performance with a model-based reward. Our VRAG-RL optimizes VLMs for RAG tasks using specially designed RL strategies, aligning the model with real-world applications. The code is available at https://github.com/Alibaba-NLP/VRAG.

Junjie Zhang, Feng Zhao, Hanqiang Liu et al.

The booming remote sensing (RS) technology is giving rise to a novel multimodality generalization task, which requires the model to overcome data heterogeneity while possessing powerful cross-scene generalization ability. Moreover, most vision-language models (VLMs) usually describe surface materials in RS images using universal texts, lacking proprietary linguistic prior knowledge specific to different RS vision modalities. In this work, we formalize RS multimodality generalization (RSMG) as a learning paradigm, and propose a frequency-aware vision-language multimodality generalization network (FVMGN) for RS image classification. Specifically, a diffusion-based training-test-time augmentation (DTAug) strategy is designed to reconstruct multimodal land-cover distributions, enriching input information for FVMGN. Following that, to overcome multimodal heterogeneity, a multimodal wavelet disentanglement (MWDis) module is developed to learn cross-domain invariant features by resampling low and high frequency components in the frequency domain. Considering the characteristics of RS vision modalities, shared and proprietary class texts is designed as linguistic inputs for the transformer-based text encoder to extract diverse text features. For multimodal vision inputs, a spatial-frequency-aware image encoder (SFIE) is constructed to realize local-global feature reconstruction and representation. Finally, a multiscale spatial-frequency feature alignment (MSFFA) module is suggested to construct a unified semantic space, ensuring refined multiscale alignment of different text and vision features in spatial and frequency domains. Extensive experiments show that FVMGN has the excellent multimodality generalization ability compared with state-of-the-art (SOTA) methods.

Yu Li, Xiaoran Shang, Qizhi Pei et al.

Post-training data plays a pivotal role in shaping the capabilities of Large Language Models (LLMs), yet datasets are often treated as isolated artifacts, overlooking the systemic connections that underlie their evolution. To disentangle these complex relationships, we introduce the concept of \textbf{data lineage} to the LLM ecosystem and propose an automated multi-agent framework to reconstruct the evolutionary graph of dataset development. Through large-scale lineage analysis, we characterize domain-specific structural patterns, such as vertical refinement in math-oriented datasets and horizontal aggregation in general-domain corpora. Moreover, we uncover pervasive systemic issues, including \textit{structural redundancy} induced by implicit dataset intersections and the \textit{propagation of benchmark contamination} along lineage paths. To demonstrate the practical value of lineage analysis for data construction, we leverage the reconstructed lineage graph to create a \textit{lineage-aware diversity-oriented dataset}. By anchoring instruction sampling at upstream root sources, this approach mitigates downstream homogenization and hidden redundancy, yielding a more diverse post-training corpus. We further highlight lineage-centric analysis as an efficient and robust topological alternative to sample-level dataset comparison for large-scale data ecosystems. By grounding data construction in explicit lineage structures, our work advances post-training data curation toward a more systematic and controllable paradigm.

Deyi Ji, Siqi Gao, Mingyuan Tao et al.

Change Detection (CD) has been attracting extensive interests with the availability of bi-temporal datasets. However, due to the huge cost of multi-temporal images acquisition and labeling, existing change detection datasets are small in quantity, short in temporal, and low in practicability. Therefore, a large-scale practical-oriented dataset covering wide temporal phases is urgently needed to facilitate the community. To this end, the ChangeNet dataset is presented especially for multi-temporal change detection, along with the new task of "Asymmetric Change Detection". Specifically, ChangeNet consists of 31,000 multi-temporal images pairs, a wide range of complex scenes from 100 cities, and 6 pixel-level annotated categories, which is far superior to all the existing change detection datasets including LEVIR-CD, WHU Building CD, etc.. In addition, ChangeNet contains amounts of real-world perspective distortions in different temporal phases on the same areas, which is able to promote the practical application of change detection algorithms. The ChangeNet dataset is suitable for both binary change detection (BCD) and semantic change detection (SCD) tasks. Accordingly, we benchmark the ChangeNet dataset on six BCD methods and two SCD methods, and extensive experiments demonstrate its challenges and great significance. The dataset is available at https://github.com/jankyee/ChangeNet.

Hongyin Zhang, Diyuan Shi, Zifeng Zhuang et al.

Developing robotic intelligent systems that can adapt quickly to unseen wild situations is one of the critical challenges in pursuing autonomous robotics. Although some impressive progress has been made in walking stability and skill learning in the field of legged robots, their ability for fast adaptation is still inferior to that of animals in nature. Animals are born with a massive set of skills needed to survive, and can quickly acquire new ones, by composing fundamental skills with limited experience. Inspired by this, we propose a novel framework, named Robot Skill Graph (RSG) for organizing a massive set of fundamental skills of robots and dexterously reusing them for fast adaptation. Bearing a structure similar to the Knowledge Graph (KG), RSG is composed of massive dynamic behavioral skills instead of static knowledge in KG and enables discovering implicit relations that exist in between the learning context and acquired skills of robots, serving as a starting point for understanding subtle patterns existing in robots' skill learning. Extensive experimental results demonstrate that RSG can provide reliable skill inference upon new tasks and environments, and enable quadruped robots to adapt to new scenarios and quickly learn new skills.

Feiyu Jia, Xiaojie Niu, Sizhe Yang et al.

Teleoperation is a key approach for collecting high-quality, physically consistent demonstrations for robotic manipulation. However, teleoperation for dexterous manipulation remains constrained by: (i) inaccurate hand-robot motion mapping, which limits teleoperated dexterity, and (ii) limited tactile feedback that forces vision-dominated interaction and hinders perception of contact geometry and force variation. To address these challenges, we present TAG, a low-cost glove system that integrates precise hand motion capture with high-resolution tactile feedback, enabling effective tactile-in-the-loop dexterous teleoperation. For motion capture, TAG employs a non-contact magnetic sensing design that provides drift-free, electromagnetically robust 21-DoF joint tracking with joint angle estimation errors below 1 degree. Meanwhile, to restore tactile sensation, TAG equips each finger with a 32-actuator tactile array within a compact 2 cm^2 module, allowing operators to directly feel physical interactions at the robot end-effector through spatial activation patterns. Through real-world teleoperation experiments and user studies, we show that TAG enables reliable real-time perception of contact geometry and dynamic force, improves success rates in contact-rich teleoperation tasks, and increases the reliability of demonstration data collection for learning-based manipulation.

Zizheng Yang, Hu Yu, Bing Li et al.

Diffusion models have recently been investigated as powerful generative solvers for image dehazing, owing to their remarkable capability to model the data distribution. However, the massive computational burden imposed by the retraining of diffusion models, coupled with the extensive sampling steps during the inference, limit the broader application of diffusion models in image dehazing. To address these issues, we explore the properties of hazy images in the semantic latent space of frozen pre-trained diffusion models, and propose a Diffusion Latent Inspired network for Image Dehazing, dubbed DiffLI$^2$D. Specifically, we first reveal that the semantic latent space of pre-trained diffusion models can represent the content and haze characteristics of hazy images, as the diffusion time-step changes. Building upon this insight, we integrate the diffusion latent representations at different time-steps into a delicately designed dehazing network to provide instructions for image dehazing. Our DiffLI$^2$D avoids re-training diffusion models and iterative sampling process by effectively utilizing the informative representations derived from the pre-trained diffusion models, which also offers a novel perspective for introducing diffusion models to image dehazing. Extensive experiments on multiple datasets demonstrate that the proposed method achieves superior performance to existing image dehazing methods. Code is available at https://github.com/aaaasan111/difflid.

Feng Zhao, Kangzheng Liu, Teng Peng et al.

Accurate representation of multimodal knowledge is crucial for event forecasting in real-world scenarios. However, existing studies have largely focused on static settings, overlooking the dynamic acquisition and fusion of multimodal knowledge. 1) At the knowledge acquisition level, how to learn time-sensitive information of different modalities, especially the dynamic structural modality. Existing dynamic learning methods are often limited to shallow structures across heterogeneous spaces or simple unispaces, making it difficult to capture deep relation-aware geometric features. 2) At the knowledge fusion level, how to learn evolving multimodal fusion features. Existing knowledge fusion methods based on static coattention struggle to capture the varying historical contributions of different modalities to future events. To this end, we propose DyMRL, a Dynamic Multispace Representation Learning approach to efficiently acquire and fuse multimodal temporal knowledge. 1) For the former issue, DyMRL integrates time-specific structural features from Euclidean, hyperbolic, and complex spaces into a relational message-passing framework to learn deep representations, reflecting human intelligences in associative thinking, high-order abstracting, and logical reasoning. Pretrained models endow DyMRL with time-sensitive visual and linguistic intelligences. 2) For the latter concern, DyMRL incorporates advanced dual fusion-evolution attention mechanisms that assign dynamic learning emphases equally to different modalities at different timestamps in a symmetric manner. To evaluate DyMRL's event forecasting performance through leveraging its learned multimodal temporal knowledge in history, we construct four multimodal temporal knowledge graph benchmarks. Extensive experiments demonstrate that DyMRL outperforms state-of-the-art dynamic unimodal and static multimodal baseline methods.

Xiaoxiao Ma, Jiachen Lei, Tianfei Ren et al.

Reinforcement learning (RL) has been successfully applied to autoregressive (AR) and diffusion models. However, extending RL to hybrid AR-diffusion frameworks remains challenging due to interleaved inference and noisy log-probability estimation. In this work, we study masked autoregressive models (MAR) and show that the diffusion head plays a critical role in training dynamics, often introducing noisy gradients that lead to instability and early performance saturation. To address this issue, we propose a stabilized RL framework for MAR. We introduce multi-trajectory expectation (MTE), which estimates the optimization direction by averaging over multiple diffusion trajectories, thereby reducing diffusion-induced gradient noise. To avoid over-smoothing, we further estimate token-wise uncertainty from multiple trajectories and apply multi-trajectory optimization only to the top-k% uncertain tokens. In addition, we introduce a consistency-aware token selection strategy that filters out AR tokens that are less aligned with the final generated content. Extensive experiments across multiple benchmarks demonstrate that our method consistently improves visual quality, training stability, and spatial structure understanding over baseline GRPO and pre-RL models. Code is available at: https://github.com/AMAP-ML/mar-grpo.

Hang Xu, Linjiang Huang, Feng Zhao

Test-time scaling (TTS) aims to achieve better results by increasing random sampling and evaluating samples based on rules and metrics. However, in text-to-image(T2I) diffusion models, most related works focus on search strategies and reward models, yet the impact of the stochastic characteristic of noise in T2I diffusion models on the method's performance remains unexplored. In this work, we analyze the effects of randomness in T2I diffusion models and explore a new format of randomness for TTS: text embedding perturbation, which couples with existing randomness like SDE-injected noise to enhance generative diversity and quality. We start with a frequency-domain analysis of these formats of randomness and their impact on generation, and find that these two randomness exhibit complementary behavior in the frequency domain: spatial noise favors low-frequency components (early steps), while text embedding perturbation enhances high-frequency details (later steps), thereby compensating for the potential limitations of spatial noise randomness in high-frequency manipulation. Concurrently, text embedding demonstrates varying levels of tolerance to perturbation across different dimensions of the generation process. Specifically, our method consists of two key designs: (1) Introducing step-based text embedding perturbation, combining frequency-guided noise schedules with spatial noise perturbation. (2) Adapting the perturbation intensity selectively based on their frequency-specific contributions to generation and tolerance to perturbation. Our approach can be seamlessly integrated into existing TTS methods and demonstrates significant improvements on multiple benchmarks with almost no additional computation. Code is available at \href{https://github.com/xuhang07/TEP-Diffusion}{https://github.com/xuhang07/TEP-Diffusion}.

Shun Zou, Yong Wang, Zehui Chen et al.

Diffusion Large Language Models (dLLMs) have recently become a promising alternative to autoregressive large language models (ARMs). Semi-autoregressive (Semi-AR) decoding is widely employed in base dLLMs and advanced decoding strategies due to its superior performance. However, our observations reveal that Semi-AR decoding suffers from inherent block constraints, which cause the decoding of many cross-block stable tokens to be unnecessarily delayed. To address this challenge, we systematically investigate the identification of stable tokens and present three key findings: (1) naive lookahead decoding is unreliable, (2) token stability closely correlates with convergence trend, and (3) historical information is isolated. Building on these insights, we propose Anchor-based History-stable Decoding (AHD), a training-free, plug-and-play dynamic decoding strategy. Specifically, AHD monitors the stability trend of tokens in real time through dynamic anchors. Once a token reaches stability, it initiates early cross-block decoding to enhance efficiency and performance. Extensive experiments across language, vision-language, and audio-language domains demonstrate that AHD simultaneously improves both performance and inference efficiency. Notably, AHD effectively reverses the performance degradation typically observed in existing advanced decoding acceleration strategies. For instance, on the BBH benchmark, our approach reduces decoding steps by 80% while improving performance by 3.67%.

Jiangtong Tan, Hu Yu, Jie Huang et al.

Long video generation involves generating extended videos using models trained on short videos, suffering from distribution shifts due to varying frame counts. It necessitates the use of local information from the original short frames to enhance visual and motion quality, and global information from the entire long frames to ensure appearance consistency. Existing training-free methods struggle to effectively integrate the benefits of both, as appearance and motion in videos are closely coupled, leading to motion inconsistency and visual quality. In this paper, we reveal that global and local information can be precisely decoupled into consistent appearance and motion intensity information by applying Principal Component Analysis (PCA), allowing for refined complementary integration of global consistency and local quality. With this insight, we propose FreePCA, a training-free long video generation paradigm based on PCA that simultaneously achieves high consistency and quality. Concretely, we decouple consistent appearance and motion intensity features by measuring cosine similarity in the principal component space. Critically, we progressively integrate these features to preserve original quality and ensure smooth transitions, while further enhancing consistency by reusing the mean statistics of the initial noise. Experiments demonstrate that FreePCA can be applied to various video diffusion models without requiring training, leading to substantial improvements. Code is available at https://github.com/JosephTiTan/FreePCA.