Recent multi-view diffusion (MVD) methods have utilized the generative capabilities of 2D image diffusion models to produce multi-view images from a single-view input. However, existing approaches often depend on dense crossview attention layers, which hinder scalability and fidelity due to their high computational costs. In this paper, we propose CTR3D, a novel method that incorporates token reduction in multi-view attention layers to efficiently generate dense, high-resolution multi-view images without restricting the camera viewpoints of the generated views. Our approach is designed into three key steps: redundancy removal, attention interaction, and token recovery. These steps leverage lightweight, projection-based techniques for multi-view token reduction and recovery, significantly improving the computational efficiency of MVD. By reducing the number of tokens in attention layers while preserving multi-view consistency, our model achieves state-of-the-art performance in novel view synthesis and 3D reconstruction while keeping efficiency for generation of dense high-resolution images and normals. Experimental results demonstrate that our method surpasses existing approaches, providing a more efficient and effective solution for multi-view generation. https://github.com/HKUST-SAIL/CTR3D

Chain-of-Thought (CoT) reasoning has been extensively explored in large models to tackle complex understanding tasks. However, it remains an open question whether such strategies can be applied to verifying and reinforcing image generation scenarios. In this paper, we provide the first comprehensive investigation in the potential of CoT reasoning to enhance autoregressive image generation. We focus on three techniques: scaling test-time computation for verification, aligning model preferences with Direct Preference Optimization (DPO), and integrating these techniques for complementary effects. Our results demonstrate that these approaches can be effectively adapted and combined to significantly improve image generation performance. Furthermore, given the pivotal role of reward models in our findings, we propose the Potential Assessment Reward Model (PARM) specialized for autoregressive image generation. PARM adaptively assesses each generation step through a potential assessment mechanism, merging the strengths of existing reward models. Using our investigated reasoning strategies, we enhance a baseline model, Show-o, to achieve superior results, with a significant +24% improvement on the GenEval benchmark, surpassing Stable Diffusion 3 by +15%. We hope our study provides unique insights and paves a new path for integrating CoT reasoning with autoregressive image generation. Code is released at https://github.com/ZiyuGuo99/Image-Generation-CoT.

Yuxuan Hu, Jihao Liu, Ke Wang, Jinliang Zheng, Weikang Shi, Manyuan Zhang, Qi Dou, Rui Liu, Aojun Zhou, Hongsheng Li. Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing. 2025.

FlowFormer [24] introduces a transformer architecture into optical flow estimation and achieves state-of-the-art performance. The core component of FlowFormer is the transformer-based cost-volume encoder. Inspired by the recent success of masked autoencoding (MAE) pretraining in unleashing transformers' capacity of encoding visual representation, we propose Masked Cost Volume Autoencoding (MCVA) to enhance FlowFormer by pretraining the cost-volume encoder with a novel MAE scheme. Firstly, we introduce a block-sharing masking strategy to prevent masked information leakage, as the cost maps of neighboring source pixels are highly correlated. Secondly, we propose a novel pre-text reconstruction task, which encourages the cost-volume encoder to aggregate long-range information and ensures pretraining-finetuning consistency. We also show how to modify the FlowFormer architecture to accommodate masks during pretraining. Pretrained with MCVA, FlowFormer++ ranks 1st among published methods on both Sintel and KITTI-2015 benchmarks. Specifically, FlowFormer++ achieves 1.07 and 1.94 average end-point error (AEPE) on the clean and final pass of Sintel benchmark, leading to 7.76% and 7.18% error reductions from FlowFormer. FlowFormer++ obtains 4.52 F1-all on the KITTI-2015 test set, improving FlowFormer by 0.16.

We introduce VideoFlow, a novel optical flow estimation framework for videos. In contrast to previous methods that learn to estimate optical flow from two frames, VideoFlow concurrently estimates bi-directional optical flows for multiple frames that are available in videos by sufficiently exploiting temporal cues.We first propose a TRi-frame Optical Flow (TROF) module that estimates bi-directional optical flows for the center frame in a three-frame manner. The information of the frame triplet is iteratively fused onto the center frame. To extend TROF for handling more frames, we further propose a MOtion Propagation (MOP) module that bridges multiple TROFs and propagates motion features between adjacent TROFs. With the iterative flow estimation refinement, the information fused in individual TROFs can be propagated into the whole sequence via MOP. By effectively exploiting video information, VideoFlow presents extraordinary performance, ranking 1st on all public benchmarks. On the Sintel benchmark, VideoFlow achieves 1.649 and 0.991 average end-point-error (AEPE) on the final and clean passes, a 15.1% and 7.6% error reduction from the best published results (1.943 and 1.073 from FlowFormer++). On the KITTI-2015 benchmark, VideoFlow achieves an F1-all error of 3.65%, a 19.2% error reduction from the best published result (4.52% from FlowFormer++). Code is released at https://github.com/XiaoyuShi97/VideoFlow.

The introduction of DETR represents a new paradigm for object detection. However, its decoder conducts classification and box localization using shared queries and cross-attention layers, leading to suboptimal results. We observe that different regions of interest in the visual feature map are suitable for performing query classification and box localization tasks, even for the same object. Salient regions provide vital information for classification, while the boundaries around them are more favorable for box regression. Unfortunately, such spatial misalignment between these two tasks greatly hinders DETR’s training. Therefore, in this work, we focus on decoupling localization and classification tasks in DETR. To achieve this, we introduce a new design scheme called spatially decoupled DETR (SD-DETR), which includes a task-aware query generation module and a disentangled feature learning process. We elaborately design the task-aware query initialization process and divide the cross-attention block in the decoder to allow the task-aware queries to match different visual regions. Meanwhile, we also observe that the prediction misalignment problem for high classification confidence and precise localization exists, so we propose an alignment loss to further guide the spatially decoupled DETR training. Through extensive experiments, we demonstrate that our approach achieves a significant improvement in MSCOCO datasets compared to previous work. For instance, we improve the performance of Conditional DETR by 4.5 AP. By spatially disentangling the two tasks, our method overcomes the misalignment problem and greatly improves the performance of DETR for object detection.

Optimizing the K-class hyperplanes in the latent space has become the standard paradigm for efficient representation learning. However, it’s almost impossible to find an optimal K-class hyperplane to accurately describe the latent space of massive noisy data. For this potential problem, we constructively propose a new method, named Switchable K-class Hyperplanes (SKH), to sufficiently describe the latent space by the mixture of K-class hyperplanes. It can directly replace the conventional single K-class hyperplane optimization as the new paradigm for noise-robust representation learning. When collaborated with the popular ArcFace on million-level data representation learning, we found that the switchable manner in SKH can effectively eliminate the gradient conflict generated by real-world label noise on a single K-class hyperplane. Moreover, combined with the margin-based loss functions (e.g. ArcFace), we propose a simple Posterior Data Clean strategy to reduce the model optimization deviation on clean dataset caused by the reduction of valid categories in each K-class hyperplane. Extensive experiments demonstrate that the proposed SKH easily achieves new state-of-the-art on IJB-B and IJB-C by encouraging noise-robust representation learning. Our code will be available at https://github.com/liubx07/SKH.git.

Radio-frequency (RF) tomographic imaging is a promising technique for inferring multi-dimensional physical space by processing RF signals traversed across a region of interest. However, conventional RF tomography schemes are generally based on vector compressed sensing, which ignores the geometric structures of the target spaces and leads to low recovery precision. The recently proposed transform-based tensor model is more appropriate for sensory data processing, as it helps exploit the geometric structures of the three-dimensional target and improve the recovery precision. In this paper, we propose a novel tensor sensing approach that achieves highly accurate estimation for real-world three-dimensional spaces. First, we use the transform-based tensor model to formulate a tensor sensing problem, and propose a fast alternating minimization algorithm called Alt-Min. Secondly, we drive an algorithm which is optimized to reduce memory and computation requirements. Finally, we present evaluation of our Alt-Min approach using IKEA 3D data and demonstrate significant improvement in recovery error and convergence speed compared to prior tensor-based compressed sensing.