Multi-arm collaboration is becoming a core capability in embodied manipulation. Recent vision-language-action (VLA) models integrate perception, language, and control, but most represent language as a single global instruction and do not provide an explicit mechanism for assigning and composing arm-specific behaviors. This design limits transfer to collaboration patterns that differ from those observed during training. We present MA-VLA, a unified framework for multi-arm collaboration via atomic action assignment. MA-VLA decomposes cooperative behavior into mid-level atomic prompts and allocates them to individual arms, enabling explicit subgoal specification and compositional reuse across tasks. To reduce reliance on fixed execution roles, we introduce Arm Shuffle, a training-time permutation of the observation, state, and assigned atomic prompts for each arm. This permutation enforces role-agnostic instruction following and supports recomposition into unseen coordination patterns, which we term multi-arm compositional generalization. We also construct a benchmark in which test-time collaboration patterns are absent in training set. Across simulation and real-world evaluations, prior state-of-the-art VLAs largely fail under these unseen collaborations, while MA-VLA consistently succeeds. These results indicate that structured, per-arm atomic action assignment offers a practical route to scalable generalization in multi-arm embodied systems. Code, models, and data are available at https://github.com/zhangzaibin/future-robots

The rapid advancement of image generation and editing techniques has rendered the detection and precise localization of forged content an increasingly demanding task. Existing approaches predominantly rely on learning forgery-specific artifacts, which limits their ability to generalize to unseen manipulation types. In this work, we introduce a Generalizable Image Forgery Localization (GIFL) framework that reframes the problem: rather than seeking manipulation traces, we propose to model the intrinsic distribution of authentic image content. GIFL learns a universal, content-consistent representation from pristine regions, organizes the feature space to naturally separate manipulated areas, and constructs a cohesive representation of authenticity that generalizes across diverse forgery types. To further improve robustness, we design a dual-domain interaction module that integrates complementary spectral and spatial cues for reliable localization. Additionally, to advance research on forgeries produced by modern deep generative models, we present Forgery ADE, a new comprehensive dataset containing images edited with a variety of popular deep image editing methods. Extensive experiments demonstrate that our method outperforms existing methods in localizing unseen forgeries also demonstrates competitive results on trained manipulation types, offering a more practical and robust solution for image authenticity verification in the era of generative AI.

Deep image inpainting techniques produce visually seamless forgeries that pose unique challenges to conventional forgery detectors due to their local coherence and semantic consistency. To address this, we propose a Fine-Grained Graph Convolution Network (IFL-GCN) for Inpainting Forgery Localization. Departing from prior local-global forgery detectors, IFL-GCN introduces a pixel-wise graph construction that achieves direct integration of local forgery traces across the entire image. By modeling each pixel as a graph node, our approach captures long-range inconsistencies while preserving fine-grained detail perception, enabling comprehensive analysis of irregular artifact distributions at the global scale. Additionally, we develop a Fidelity-aware Weighted Loss (FW loss) to dynamically calibrate learning objectives based on the estimated fidelity of the forged content, enhancing the detector's sensitivity to subtle, high-fidelity forgeries. Furthermore, to bridge the generalization gap across diverse inpainting artifacts and improve robustness, we propose Forgery Intensity Mixup augmentation (FIM), which expands the distribution of forgery features by modulating the intensity of forgeries while maintaining spatial-semantic integrity. Extensive experiments across 6 mainstream forgery detection benchmarks demonstrate that IFL-GCN achieves state-of-the-art performance, outperforming the closest competing method by 6.7% in average F1 score over all inpainting forgery test sets.

While Vision-Language Models (VLMs) excel at 2D visual understanding, they remain constrained by 2D-centric paradigm that severely limits genuine 3D spatial reasoning. To bridge this gap, we introduce Think3D, a novel framework that equips VLM agents with interactive, 3D chain-of-thought reasoning capabilities. By integrating a suite of 3D manipulation tools, Think3D transforms perception into active spatial exploration, mirroring human geometric reasoning. Think3D consistently improves proprietary models, including GPT-4.1 and Gemini 2.5 Pro, across BLINK Multi-view, MindCube-1K, and VSI-Bench-Tiny. We further propose Think3D-RL to teach smaller open-weight models how to manipulate 3D space effectively. Using only final-answer rewards, without process supervision or handcrafted exploration trajectories, Think3D-RL enables Qwen3-VL-4B to autonomously learn effective 3D exploration strategies. After training, the model exhibits tool-use patterns similar to those of stronger proprietary models, while shifting the effect of 3D tool use on MindCube-1K from a performance drop to a substantial improvement. These results show that active exploration in 3D space provides an effective and general paradigm for improving spatial reasoning in multimodal agents. Code, models, and data are available at https://github.com/zhangzaibin/spagent.

Salient object detection is inherently a subjective problem, as observers with different priors may perceive different objects as salient. However, existing methods predominantly formulate it as an objective prediction task with a single groundtruth segmentation map for each image, which renders the problem under-determined and fundamentally ill-posed. To address this issue, we propose Observer-Centric Salient Object Detection (OC-SOD), where salient regions are predicted by considering not only the visual cues but also the observer-specific factors such as their preferences or intents. As a result, this formulation captures the intrinsic ambiguity and diversity of human perception, enabling personalized and context-aware saliency prediction. By leveraging multi-modal large language models, we develop an efficient data annotation pipeline and construct the first OC-SOD dataset named OC-SODBench, comprising 33k training, validation and test images with 152k textual prompts and object pairs. Built upon this new dataset, we further design OC-SODAgent, an agentic baseline which performs OC-SOD via a human-like "Perceive-Reflect-Adjust" process. Extensive experiments on our proposed OC-SODBench have justified the effectiveness of our contribution. Through this observer-centric perspective, we aim to bridge the gap between human perception and computational modeling, offering a more realistic and flexible understanding of what makes an object truly "salient." Code and dataset are publicly available at: https://github.com/Dustzx/OC_SOD

Achieving fully autonomous driving systems requires learning rational decisions in a wide span of scenarios, including safety-critical and out-of-distribution ones. However, such cases are underrepresented in real-world corpus collected by human experts. To complement for the lack of data diversity, we introduce a novel and scalable simulation framework capable of synthesizing these crucial massive unseen states upon existing driving logs. Our pipeline utilizes advanced neural rendering with a reactive environment to generate high-fidelity multi-view observations controlled by ego trajectory perturbations. Furthermore, we develop a pseudo-expert trajectory generation mechanism to provide feasible action supervision for these newly simulated states to provide action supervision.Upon the synthesized data, we find that a simple co-training strategy on both real-world and simulated samples can lead to significant improvements in both robustness and generalization for various planning methods on challenging real-world benchmarks, up to +6.8 EPDMS on navhard and +2.9 on navtest. More importantly, such policy improvement scales smoothly by increasing simulation data only, even without extra real-world data streaming in. We further reveal crucial findings of such a sim-real paradigm, includingthe design of pseudo-experts and the scaling properties for different policy architectures. Simulation data and code would be released.

As multi-object tracking (MOT) tasks continue to evolve toward more general and multi-modal scenarios, the rigid and task-specific architectures of existing MOT methods increasingly hinder their applicability across diverse tasks and limit flexibility in adapting to new tracking formulations. Most approaches rely on fixed output heads and bespoke tracking pipelines, making them difficult to extend to more complex or instruction-driven tasks. To address these limitations, we propose AR-MOT, a novel autoregressive paradigm that formulates MOT as a sequence generation task within a large language model (LLM) framework. This design enables the model to output structured results through flexible sequence construction, without requiring any task-specific heads. To enhance region-level visual perception, we introduce an Object Tokenizer based on a pretrained detector. To mitigate the misalignment between global and regional features, we propose a Region-Aware Alignment (RAA) module, and to support long-term tracking, we design a Temporal Memory Fusion (TMF) module that caches historical object tokens. AR-MOT offers strong potential for extensibility, as new modalities or instructions can be integrated by simply modifying the output sequence format without altering the model architecture. Extensive experiments on MOT17 and DanceTrack validate the feasibility of our approach, achieving performance comparable to state-of-the-art methods while laying the foundation for more general and flexible MOT systems.

Existing self-supervised methods for depth-pose joint learning mainly focus on the design of sophisticated depth estimation networks, while pose estimation is mostly conducted using a vanilla black-box CNN-based PoseNet that directly regresses the relative pose from two frames. Since direct pose regression is a non-trivial task requiring abstract understanding of the scenes, self-supervised PoseNet can hardly generalize to challenging scenarios with complex camera motions, leading to inferior performance of both pose and depth estimation. To address this issue, we propose a pose estimation method that explicitly recovers relative poses from 3D-2D correspondences, in conjunction with depth estimation via a differentiable weighted Perspective-n-Point (PnP) module. Global pose estimation from input images then essentially amounts to local image matching, which mainly relies on low-level cues and is therefore more concrete with stronger generalization ability. In addition, a weighting module is learned to adaptively filter out outliers based on their appearance feature, since not all matches contribute equally in solving relative poses, and only reliable ones are needed for robust pose recovery. We implement the above pipeline using a fully differentiable network, which allows self-supervised joint depth-pose learning in an end-to-end manner. Extensive experiments show that our method achieves stateof-the-art performance on KITTI monocular depth estimation and visual odometry benchmarks, with a 2%-5% improvement in depth accuracy (Abs Rel) using the identical model architecture and 40% improvement in pose accuracy (ATE) compared with PoseNet. Furthermore, our approach demonstrates superior generalization ability on indoor scenarios, outperforming existing self-supervised counterparts on the NYUv2 dataset by 1.5% in depth accuracy (Abs Rel).

Recent single-image to 3D generation methods commonly adopt multi-view diffusion and large reconstruction models to achieve fast 3D content generation. Despite the impressive generation speed and geometrical consistency, there are multiple deficiencies including texture distortion, color deviation, and insufficient resolution. To address these deficiencies, we present DreamTexture, a high-fidelity method for decoupling geometry and texture synthesis in two stages. The main idea is to combine the powerful geometrical generation capability of the large multi-view Gaussian model with the texture alignment ability of the synchronized multi-view diffusion strategy. To synthesize subject-driven personalized textures for decorating the 3D object, we train a personalized diffusion model to generate subject-driven multi-view images. These images are then mapped to the texture domain to synthesize high-resolution textures. To improve the quality of multi-view images for further optimizing texture details, we instantiate two scaling factors to rebalance the contributions of backbone features and skip features in personalized diffusion. Experimental results on public datasets demonstrate that DreamTexture significantly outperforms the latest state-of-the-art methods, both qualitatively and quantitatively. Notably, our framework can generate high-fidelity 3D assets with detailed textures from scratch, featuring remarkable training scalability.

The essence of audio-visual segmentation (AVS) lies in locating and delineating sound-emitting objects within a video stream. While Transformer-based methods have shown promise, their handling of long-range dependencies struggles due to quadratic computational costs, presenting a bottleneck in complex scenarios. To overcome this limitation and facilitate complex multi-modal comprehension with linear complexity, we introduce AVS-Mamba, a selective state space model to address the AVS task. Our framework incorporates two key components for video understanding and cross-modal learning: Temporal Mamba Block for sequential video processing and Vision-to-Audio Fusion Block for advanced audio-vision integration. Building on this, we develop the Multi-scale Temporal Encoder, aimed at enhancing the learning of visual features across scales, facilitating the perception of intra- and inter-frame information. To perform multi-modal fusion, we propose the Modality Aggregation Decoder, leveraging the Vision-to-Audio Fusion Block to integrate visual features into audio features across both frame and temporal levels. Further, we adopt the Contextual Integration Pyramid to perform audio-to-vision spatial-temporal context collaboration. Through these innovative contributions, our approach achieves new state-of-the-art results on the AVSBench-object and AVSBench-semantic datasets. Our source code and model weights are available at AVS-Mamba.

Text logo design heavily relies on the creativity and expertise of professional designers, in which arranging element layouts is one of the most important procedures. However, this specific task has received limited attention, often overshadowed by broader layout generation tasks such as document or poster design. In this paper, we propose a Vision-Language Model (VLM)-based framework that generates content-aware text logo layouts by integrating multi-modal inputs with user-defined constraints, enabling more flexible and robust layout generation for real-world applications. We introduce two model techniques that reduce the computational cost for processing multiple glyph images simultaneously, without compromising performance. To support instruction tuning of our model, we construct two extensive text logo datasets that are five times larger than existing public datasets. In addition to geometric annotations (e.g., text masks and character recognition), our datasets include detailed layout descriptions in natural language, enabling the model to reason more effectively in handling complex designs and custom user inputs. Experimental results demonstrate the effectiveness of our proposed framework and datasets, outperforming existing methods on various benchmarks that assess geometric aesthetics and human preferences.

Despite remarkable progress in driving world models, their potential for autonomous systems remains largely untapped: the world models are mostly learned for world simulation and decoupled from trajectory planning. While recent efforts aim to unify world modeling and planning in a single framework, the synergistic facilitation mechanism of world modeling for planning still requires further exploration. In this work, we introduce a new driving paradigm named Policy World Model (PWM), which not only integrates world modeling and trajectory planning within a unified architecture, but is also able to benefit planning using the learned world knowledge through the proposed action-free future state forecasting scheme. Through collaborative state-action prediction, PWM can mimic the human-like anticipatory perception, yielding more reliable planning performance. To facilitate the efficiency of video forecasting, we further introduce a parallel token generation mechanism, equipped with a context-guided tokenizer and an adaptive dynamic focal loss. Despite utilizing only front camera input, our method matches or exceeds state-of-the-art approaches that rely on multi-view and multi-modal inputs. Code will be released at https://github.com/6550Zhao/Policy-World-Model.

Visual Spatial Reasoning (VSR) is a core human cognitive ability and a critical requirement for advancing embodied intelligence and autonomous systems. Despite recent progress in Vision-Language Models (VLMs), achieving human-level VSR remains highly challenging due to the complexity of representing and reasoning over three-dimensional space. In this paper, we present a systematic investigation of VSR in VLMs, encompassing a review of existing methodologies across input modalities, model architectures, training strategies, and reasoning mechanisms. Furthermore, we categorize spatial intelligence into three levels of capability, ie, basic perception, spatial understanding, spatial planning, and curate SIBench, a spatial intelligence benchmark encompassing nearly 20 open-source datasets across 23 task settings. Experiments with state-of-the-art VLMs reveal a pronounced gap between perception and reasoning, as models show competence in basic perceptual tasks but consistently underperform in understanding and planning tasks, particularly in numerical estimation, multi-view reasoning, temporal dynamics, and spatial imagination. These findings underscore the substantial challenges that remain in achieving spatial intelligence, while providing both a systematic roadmap and a comprehensive benchmark to drive future research in the field. The related resources of this study are accessible at https://sibench.github.io/Awesome-Visual-Spatial-Reasoning/.

In recent years, advanced image editing and generation methods have rapidly evolved, making detecting and locating forged image content increasingly challenging. Most existing image forgery detection methods rely on identifying the edited traces left in the image. However, because the traces of different forgeries are distinct, these methods can identify familiar forgeries included in the training data but struggle to handle unseen ones. In response, we present an approach for Generalizable Image Forgery Localization (GIFL). Once trained, our model can detect both seen and unseen forgeries, providing a more practical and efficient solution to counter false information in the era of generative AI. Our method focuses on learning general features from the pristine content rather than traces of specific forgeries, which are relatively consistent across different types of forgeries and therefore can be used as universal features to locate unseen forgeries. Additionally, as existing image forgery datasets are still dominated by traditional hand-crafted forgeries, we construct a new dataset consisting of images edited by various popular deep generative image editing methods to further encourage research in detecting images manipulated by deep generative models. Extensive experimental results show that the proposed approach outperforms state-of-the-art methods in the detection of unseen forgeries and also demonstrates competitive results for seen forgeries. The code and dataset are available at https://github.com/ZhaoHengrun/GIFL.

With the rapid proliferation of 3D devices and the shortage of 3D content, stereo conversion is attracting increasing attention. Recent works introduce pretrained Diffusion Models (DMs) into this task. However, due to the scarcity of large-scale training data and comprehensive benchmarks, the optimal methodologies for employing DMs in stereo conversion and the accurate evaluation of stereo effects remain largely unexplored. In this work, we introduce the Mono2Stereo dataset, providing high-quality training data and benchmark to support in-depth exploration of stereo conversion. With this dataset, we conduct an empirical study that yields two primary findings. 1) The differences between the left and right views are subtle, yet existing metrics consider overall pixels, failing to concentrate on regions critical to stereo effects. 2) Mainstream methods adopt either one-stage left-to-right generation or warp-and-inpaint pipeline, facing challenges of degraded stereo effect and image distortion respectively. Based on these findings, we introduce a new evaluation metric, Stereo Intersection-over-Union, which prioritizes disparity and achieves a high correlation with human judgments on stereo effect. Moreover, we propose a strong baseline model, harmonizing the stereo effect and image quality simultaneously, and notably surpassing current mainstream methods. Our code and data will be open-sourced to promote further research in stereo conversion. Our models are available at mono2stereo-bench.github.io.

Recently, unsupervised domain adaptive object detection methods have been proposed to address the challenge of detecting objects across different domains without labeled data in the target domain. These methods focus on aligning features either at the image level or the instance level. However, due to the absence of annotations in the target domain, existing approaches encounter challenges such as background noise at the image level and prototype aggregation noise at the instance level. To tackle these issues, we introduce a novel adaptive feature alignment network for cross-domain object detection, comprising two key modules. Firstly, we present an adaptive foreground-aware attention module equipped with a set of learnable part prototypes for image-level alignment. This module dynamically generates foreground attention maps, enabling the model to prioritize foreground features, thus reducing the impact of background noise. Secondly, we propose a class- aware prototype alignment module incorporating an optimal transport algorithm for instance-level alignment. This module mitigates the adverse effects of region-prototype aggregation noise by aligning prototypes with instances based on their semantic similarities. By integrating these two modules, our approach achieves better image-level and instance-level feature alignment. Extensive experiments across three challenging scenarios demonstrate the effectiveness of our method, outperforming state-of-the-art approaches in terms of object detection performance.

Audio-Visual Segmentation (AVS) aims to segment sound-producing objects in videos according to associated audio cues, where both modalities are affected by noise to different extents, such as the blending of background noises in audio or the presence of distracted objects in video. Most existing methods focus on learning interactions between modalities at high semantic levels but is incapable of filtering low-level noise or achieving fine-grained representational interactions during the early feature extraction phase. Consequently, they struggle with illusion issues, where nonexistent audio cues are erroneously linked to visual objects. In this paper, we present SelM, a novel architecture that leverages selective mechanisms to counteract these illusions. SelM employs State Space model for noise reduction and robust feature selection. By imposing additional bidirectional constraints on audio and visual embeddings, it is able to precisely identify crucial features corresponding to sound-emitting targets. To fill the existing gap in early fusion within AVS, SelM introduces a dual alignment mechanism specifically engineered to facilitate intricate spatio-temporal interactions between audio and visual streams, achieving more fine-grained representations. Moreover, we develop a cross-level decoder for layered reasoning, significantly enhancing segmentation precision by exploring the complex relationships between audio and visual information. SelM achieves state-of-the-art performance in AVS tasks, especially in the challenging Audio-Visual Semantic Segmentation subset. The code can be found at https://github.com/Cyyzpoi/SelM.

Multi-object tracking typically concentrates on the tracking of individual targets without taking into account the scenario where some of these targets are part of a formation. In this paper, we define a new task called multi-object formation tracking, which requires tracking the formation as a whole while simultaneously tracking individual targets within the formation. To tackle this challenge, we propose a new Multi-Object Formation Tracking method called MOFTrack, which is composed of three main components: (1) position search, which adaptively assigns different search ranges for targets of varying sizes, (2) trajectory validation, which addresses the issue of formation instability during target intersections, and (3) effective bit motion selection, which facilitates the rapid formation tracking of newly emerging targets while also mitigating the impact of trajectory loss on the collection of motion feature vectors. To measure the accuracy of formation tracking, we introduce a novel simulated aircraft formation dataset with formation numbers for small targets in remote sensing videos. In addition, we introduce a metric called Formation Accuracy, FACC. We categorize targets into three groups to assess performance across various target pixel sizes: extremely small, relatively small and small. We successfully implement online reliable tracking and formation assigning for multiple targets in remote sensing videos, achieving high FACC for all three categories of targets. Source code and dataset will be made publicly available.

This paper aims to design monocular depth estimation models with better generalization abilities. To this end, we have conducted quantitative analysis and discovered two important insights. First, the Simulation Correlation phenomenon, commonly seen in long-tailed classification problems, also exists in monocular depth estimation, indicating that the imbalanced depth distribution in training data may be the cause of limited generalization ability. Second, the imbalanced and long-tail distribution of depth values extends beyond the dataset scale, and also manifests within each individual image, further exacerbating the challenge of monocular depth estimation. Motivated by the above findings, we propose the Distance-aware Multi-Expert (DME) depth estimation model. Unlike prior methods that handle different depth range indiscriminately, DME adopts a divide-and-conquer philosophy where each expert is responsible for depth estimation of regions within a specific depth range. As such, the depth distribution seen by each expert is more uniform and can be more easily predicted. A pixel-level routing module is further designed and learned to stitch the prediction of all experts into the final depth map. Experiments show that DME achieves state-of-the-art performance on both NYU-Depth v2 and KITTI, and also delivers favorable zero-shot generalization capability on unseen datasets.