Jiakui Hu

Man Yao, Jiakui Hu, Zhaokun Zhou et al.

Spiking Neural Networks (SNNs) provide an energy-efficient deep learning option due to their unique spike-based event-driven (i.e., spike-driven) paradigm. In this paper, we incorporate the spike-driven paradigm into Transformer by the proposed Spike-driven Transformer with four unique properties: 1) Event-driven, no calculation is triggered when the input of Transformer is zero; 2) Binary spike communication, all matrix multiplications associated with the spike matrix can be transformed into sparse additions; 3) Self-attention with linear complexity at both token and channel dimensions; 4) The operations between spike-form Query, Key, and Value are mask and addition. Together, there are only sparse addition operations in the Spike-driven Transformer. To this end, we design a novel Spike-Driven Self-Attention (SDSA), which exploits only mask and addition operations without any multiplication, and thus having up to $87.2\times$ lower computation energy than vanilla self-attention. Especially in SDSA, the matrix multiplication between Query, Key, and Value is designed as the mask operation. In addition, we rearrange all residual connections in the vanilla Transformer before the activation functions to ensure that all neurons transmit binary spike signals. It is shown that the Spike-driven Transformer can achieve 77.1\% top-1 accuracy on ImageNet-1K, which is the state-of-the-art result in the SNN field. The source code is available at https://github.com/BICLab/Spike-Driven-Transformer.

Man Yao, Jiakui Hu, Guangshe Zhao et al.

Spiking Neural Networks (SNNs) are well known as a promising energy-efficient alternative to conventional artificial neural networks. Subject to the preconceived impression that SNNs are sparse firing, the analysis and optimization of inherent redundancy in SNNs have been largely overlooked, thus the potential advantages of spike-based neuromorphic computing in accuracy and energy efficiency are interfered. In this work, we pose and focus on three key questions regarding the inherent redundancy in SNNs. We argue that the redundancy is induced by the spatio-temporal invariance of SNNs, which enhances the efficiency of parameter utilization but also invites lots of noise spikes. Further, we analyze the effect of spatio-temporal invariance on the spatio-temporal dynamics and spike firing of SNNs. Then, motivated by these analyses, we propose an Advance Spatial Attention (ASA) module to harness SNNs' redundancy, which can adaptively optimize their membrane potential distribution by a pair of individual spatial attention sub-modules. In this way, noise spike features are accurately regulated. Experimental results demonstrate that the proposed method can significantly drop the spike firing with better performance than state-of-the-art SNN baselines. Our code is available in \url{https://github.com/BICLab/ASA-SNN}.

Xiao Cao, Yansong Qu, Xiangzhen et al.

Mask-free video object insertion has emerged as a challenging task, requiring harmonious integration of reference objects into source videos. However, existing methods struggle when references exhibit severe stylistic domain gaps with the source scene. To overcome this, we propose \textit{\textbf{Smart-Insertion-V}}, an end-to-end \textbf{Dual-Stream} framework that concurrently conducts video insertion and image style transfer. Within this framework, the image stream synchronously guides the video generation process, while a \textbf{Closed-loop Feedback} mechanism is further incorporated to ensure robust insertion. Inevitably, integrating these diverse conditioning signals results in feature entanglement and style leakage. To tackle this issue, we design \textbf{Dual-World-View RoPE} to distinguish different signals via spatial-temporal offsets without incurring heavy training overhead. Furthermore, to facilitate spatial grounding and stylistic adaptation, we introduce a \textbf{Decoupled Guidance Module} that leverages a Vision-Language Model for semantic reasoning while preserving original temporal guidance with native text encoder. To bridge data gap for harmonious reference insertion task, we propose a data curation pipeline and will release an \textbf{open-source dataset}. Experiments demonstrate that our method can insert objects into plausible positions while achieving the most harmonious results.

Man Yao, Jiakui Hu, Tianxiang Hu et al.

Neuromorphic computing, which exploits Spiking Neural Networks (SNNs) on neuromorphic chips, is a promising energy-efficient alternative to traditional AI. CNN-based SNNs are the current mainstream of neuromorphic computing. By contrast, no neuromorphic chips are designed especially for Transformer-based SNNs, which have just emerged, and their performance is only on par with CNN-based SNNs, offering no distinct advantage. In this work, we propose a general Transformer-based SNN architecture, termed as ``Meta-SpikeFormer", whose goals are: 1) Lower-power, supports the spike-driven paradigm that there is only sparse addition in the network; 2) Versatility, handles various vision tasks; 3) High-performance, shows overwhelming performance advantages over CNN-based SNNs; 4) Meta-architecture, provides inspiration for future next-generation Transformer-based neuromorphic chip designs. Specifically, we extend the Spike-driven Transformer in \citet{yao2023spike} into a meta architecture, and explore the impact of structure, spike-driven self-attention, and skip connection on its performance. On ImageNet-1K, Meta-SpikeFormer achieves 80.0\% top-1 accuracy (55M), surpassing the current state-of-the-art (SOTA) SNN baselines (66M) by 3.7\%. This is the first direct training SNN backbone that can simultaneously supports classification, detection, and segmentation, obtaining SOTA results in SNNs. Finally, we discuss the inspiration of the meta SNN architecture for neuromorphic chip design. Source code and models are available at \url{https://github.com/BICLab/Spike-Driven-Transformer-V2}.

Man Yao, Xuerui Qiu, Tianxiang Hu et al.

The ambition of brain-inspired Spiking Neural Networks (SNNs) is to become a low-power alternative to traditional Artificial Neural Networks (ANNs). This work addresses two major challenges in realizing this vision: the performance gap between SNNs and ANNs, and the high training costs of SNNs. We identify intrinsic flaws in spiking neurons caused by binary firing mechanisms and propose a Spike Firing Approximation (SFA) method using integer training and spike-driven inference. This optimizes the spike firing pattern of spiking neurons, enhancing efficient training, reducing power consumption, improving performance, enabling easier scaling, and better utilizing neuromorphic chips. We also develop an efficient spike-driven Transformer architecture and a spike-masked autoencoder to prevent performance degradation during SNN scaling. On ImageNet-1k, we achieve state-of-the-art top-1 accuracy of 78.5\%, 79.8\%, 84.0\%, and 86.2\% with models containing 10M, 19M, 83M, and 173M parameters, respectively. For instance, the 10M model outperforms the best existing SNN by 7.2\% on ImageNet, with training time acceleration and inference energy efficiency improved by 4.5$\times$ and 3.9$\times$, respectively. We validate the effectiveness and efficiency of the proposed method across various tasks, including object detection, semantic segmentation, and neuromorphic vision tasks. This work enables SNNs to match ANN performance while maintaining the low-power advantage, marking a significant step towards SNNs as a general visual backbone. Code is available at https://github.com/BICLab/Spike-Driven-Transformer-V3.

JiaKui Hu, Jialun Liu, Liying Yang et al.

Scene-consistent video generation aims to create videos that explore 3D scenes based on a camera trajectory. Previous methods rely on video generation models with external memory for consistency, or iterative 3D reconstruction and inpainting, which accumulate errors during inference due to incorrect intermediary outputs, non-differentiable processes, and separate models. To overcome these limitations, we introduce ``geometry-as-context". It iteratively completes the following steps using an autoregressive camera-controlled video generation model: (1) estimates the geometry of the current view necessary for 3D reconstruction, and (2) simulates and restores novel view images rendered by the 3D scene. Under this multi-task framework, we develop the camera gated attention module to enhance the model's capability to effectively leverage camera poses. During the training phase, text contexts are utilized to ascertain whether geometric or RGB images should be generated. To ensure that the model can generate RGB-only outputs during inference, the geometry context is randomly dropped from the interleaved text-image-geometry training sequence. The method has been tested on scene video generation with one-direction and forth-and-back trajectories. The results show its superiority over previous approaches in maintaining scene consistency and camera control.

Shuang Zeng, Boxu Xie, Lei Zhu et al.

Deep learning has greatly advanced medical image segmentation, but its success relies heavily on fully supervised learning, which requires dense annotations that are costly and time-consuming for 3D volumetric scans. Barely-supervised learning reduces annotation burden by using only a few labeled slices per volume. Existing methods typically propagate sparse annotations to unlabeled slices through geometric continuity to generate pseudo-labels, but this strategy lacks semantic understanding, often resulting in low-quality pseudo-labels. Furthermore, medical image segmentation is inherently a pixel-level visual understanding task, where accuracy fundamentally depends on the quality of local, fine-grained visual features. Inspired by this, we propose RADA, a novel Region-Aware Dual-encoder Auxiliary learning pipeline which introduces a dual-encoder framework pre-trained on Alpha-CLIP to extract fine-grained, region-specific visual features from the original images and limited annotations. The framework combines image-level fine-grained visual features with text-level semantic guidance, providing region-aware semantic supervision that bridges image-level semantics and pixel-level segmentation. Integrated into a triple-view training framework, RADA achieves SOTA performance under extremely sparse annotation settings on LA2018, KiTS19 and LiTS, demonstrating robust generalization across diverse datasets.

Xinjie Zhang, Jintao Guo, Shanshan Zhao et al.

Recent years have seen remarkable progress in both multimodal understanding models and image generation models. Despite their respective successes, these two domains have evolved independently, leading to distinct architectural paradigms: While autoregressive-based architectures have dominated multimodal understanding, diffusion-based models have become the cornerstone of image generation. Recently, there has been growing interest in developing unified frameworks that integrate these tasks. The emergence of GPT-4o's new capabilities exemplifies this trend, highlighting the potential for unification. However, the architectural differences between the two domains pose significant challenges. To provide a clear overview of current efforts toward unification, we present a comprehensive survey aimed at guiding future research. First, we introduce the foundational concepts and recent advancements in multimodal understanding and text-to-image generation models. Next, we review existing unified models, categorizing them into three main architectural paradigms: diffusion-based, autoregressive-based, and hybrid approaches that fuse autoregressive and diffusion mechanisms. For each category, we analyze the structural designs and innovations introduced by related works. Additionally, we compile datasets and benchmarks tailored for unified models, offering resources for future exploration. Finally, we discuss the key challenges facing this nascent field, including tokenization strategy, cross-modal attention, and data. As this area is still in its early stages, we anticipate rapid advancements and will regularly update this survey. Our goal is to inspire further research and provide a valuable reference for the community. The references associated with this survey are available on GitHub (https://github.com/AIDC-AI/Awesome-Unified-Multimodal-Models).

JiaKui Hu, Lujia Jin, Zhengjian Yao et al.

This paper proposes the Degradation Classification Pre-Training (DCPT), which enables models to learn how to classify the degradation type of input images for universal image restoration pre-training. Unlike the existing self-supervised pre-training methods, DCPT utilizes the degradation type of the input image as an extremely weak supervision, which can be effortlessly obtained, even intrinsic in all image restoration datasets. DCPT comprises two primary stages. Initially, image features are extracted from the encoder. Subsequently, a lightweight decoder, such as ResNet18, is leveraged to classify the degradation type of the input image solely based on the features extracted in the first stage, without utilizing the input image. The encoder is pre-trained with a straightforward yet potent DCPT, which is used to address universal image restoration and achieve outstanding performance. Following DCPT, both convolutional neural networks (CNNs) and transformers demonstrate performance improvements, with gains of up to 2.55 dB in the 10D all-in-one restoration task and 6.53 dB in the mixed degradation scenarios. Moreover, previous self-supervised pretraining methods, such as masked image modeling, discard the decoder after pre-training, while our DCPT utilizes the pre-trained parameters more effectively. This superiority arises from the degradation classifier acquired during DCPT, which facilitates transfer learning between models of identical architecture trained on diverse degradation types. Source code and models are available at https://github.com/MILab-PKU/dcpt.

Zhengjian Yao, Jiakui Hu, Kaiwen Li et al.

Blind face restoration remains a persistent challenge due to the inherent ill-posedness of reconstructing holistic structures from severely constrained observations. Current generative approaches, while capable of synthesizing realistic textures, often suffer from information asymmetry -- the intrinsic disparity between the information-sparse low quality inputs and the information-dense high quality outputs. This imbalance leads to a one-to-many mapping, where insufficient constraints result in stochastic uncertainty and hallucinatory artifacts. To bridge this gap, we present \textbf{Pref-Restore}, a hierarchical framework that integrates discrete semantic logic with continuous texture generation to achieve deterministic, preference-aligned restoration. Our methodology fundamentally addresses this information disparity through two complementary strategies: (1) Augmenting Input Density: We employ an auto-regressive integrator to reformulate textual instructions into dense latent queries, injecting high-level semantic stability to constrain the degraded signals; (2) Pruning Output Distribution: We pioneer the integration of on-policy reinforcement learning directly into the diffusion restoration loop. By transforming human preferences into differentiable constraints, we explicitly penalize stochastic deviations, thereby sharpening the posterior distribution toward the desired high-fidelity outcomes. Extensive experiments demonstrate that Pref-Restore achieves state-of-the-art performance across synthetic and real-world benchmarks. Furthermore, empirical analysis confirms that our preference-aligned strategy significantly reduces solution entropy, establishing a robust pathway toward reliable and deterministic blind restoration.

Bojun Xiong, Jialun Liu, Jiakui Hu et al.

Physically Based Rendering (PBR) materials play a crucial role in modern graphics, enabling photorealistic rendering across diverse environment maps. Developing an effective and efficient algorithm that is capable of automatically generating high-quality PBR materials rather than RGB texture for 3D meshes can significantly streamline the 3D content creation. Most existing methods leverage pre-trained 2D diffusion models for multi-view image synthesis, which often leads to severe inconsistency between the generated textures and input 3D meshes. This paper presents TexGaussian, a novel method that uses octant-aligned 3D Gaussian Splatting for rapid PBR material generation. Specifically, we place each 3D Gaussian on the finest leaf node of the octree built from the input 3D mesh to render the multi-view images not only for the albedo map but also for roughness and metallic. Moreover, our model is trained in a regression manner instead of diffusion denoising, capable of generating the PBR material for a 3D mesh in a single feed-forward process. Extensive experiments on publicly available benchmarks demonstrate that our method synthesizes more visually pleasing PBR materials and runs faster than previous methods in both unconditional and text-conditional scenarios, exhibiting better consistency with the given geometry. Our code and trained models are available at https://3d-aigc.github.io/TexGaussian.

JiaKui Hu, Shanshan Zhao, Qing-Guo Chen et al.

This paper presents Omni-View, which extends the unified multimodal understanding and generation to 3D scenes based on multiview images, exploring the principle that "generation facilitates understanding". Consisting of understanding model, texture module, and geometry module, Omni-View jointly models scene understanding, novel view synthesis, and geometry estimation, enabling synergistic interaction between 3D scene understanding and generation tasks. By design, it leverages the spatiotemporal modeling capabilities of its texture module responsible for appearance synthesis, alongside the explicit geometric constraints provided by its dedicated geometry module, thereby enriching the model's holistic understanding of 3D scenes. Trained with a two-stage strategy, Omni-View achieves a state-of-the-art score of 55.4 on the VSI-Bench benchmark, outperforming existing specialized 3D understanding models, while simultaneously delivering strong performance in both novel view synthesis and 3D scene generation.

Xuerui Qiu, Yutao Cui, Guozhen Zhang et al.

Unified Multimodal Models struggle to bridge the fundamental gap between the abstract representations needed for visual understanding and the detailed primitives required for generation. Existing approaches typically compromise by employing decoupled encoders, stacking representation encoder atop VAEs, or utilizing discrete quantization. However, these methods often disrupt information coherence and lead to optimization conflicts. To this end, we introduce HYDRA-TOK, a representation-harmonized pure ViT in the insight that visual modeling should evolve from generation to understanding. HYDRA-TOK reformulates the standard backbone into a progressive learner that transitions from a Gen-ViT, which captures structure-preserving primitives, to a Sem-ViT for semantic encoding. Crucially, this transition is mediated by a Generation-Semantic Bottleneck (GSB), which compresses features into a low-dimensional space to filter noise for robust synthesis, then restores dimensionality to empower complex semantic comprehension. Built upon this foundation, we present HYDRA, a native unified framework integrating perception and generation within a single parameter space. Extensive experiments establish HYDRA as a new state-of-the-art. It sets a benchmark in visual reconstruction (rFID 0.08) and achieves top-tier generation performance on GenEval (0.86), DPG-Bench (86.4), and WISE (0.53), while simultaneously outperforming previous native UMMs by an average of 10.0 points across eight challenging understanding benchmarks.

Jialun Liu, Yukuo Ma, Xiao Cao et al.

Recent advances in diffusion-based video generation have substantially improved visual fidelity and temporal coherence. However, most existing approaches remain task-specific and rely primarily on textual instructions, limiting their ability to handle multimodal inputs, contextual references, and diverse video generation and editing scenarios within a unified framework. Moreover, many video editing methods depend on carefully engineered pipelines tailored to individual operations, which hinders scalability and composability. In this paper, we propose Tele-Omni, a unified multimodal framework for video generation and editing that follows multimodal instructions, including text, images, and reference videos, within a single model. Tele-Omni leverages pretrained multimodal large language models to parse heterogeneous instructions and infer structured generation or editing intents, while diffusion-based generators perform high-quality video synthesis conditioned on these structured signals. To enable joint training across heterogeneous video tasks, we introduce a task-aware data processing pipeline that unifies multimodal inputs into a structured instruction format while preserving task-specific constraints. Tele-Omni supports a wide range of video-centric tasks, including text-to-video generation, image-to-video generation, first-last-frame video generation, in-context video generation, and in-context video editing. By decoupling instruction parsing from video synthesis and combining it with task-aware data design, Tele-Omni achieves flexible multimodal control while maintaining strong temporal coherence and visual consistency. Experimental results demonstrate that Tele-Omni achieves competitive performance across multiple tasks.

JiaKui Hu, Yuxiao Yang, Jialun Liu et al.

Generating multi-view images from human instructions is crucial for 3D content creation. The primary challenges involve maintaining consistency across multiple views and effectively synthesizing shapes and textures under diverse conditions. In this paper, we propose the Multi-View Auto-Regressive (\textbf{MV-AR}) method, which leverages an auto-regressive model to progressively generate consistent multi-view images from arbitrary prompts. Firstly, the next-token-prediction capability of the AR model significantly enhances its effectiveness in facilitating progressive multi-view synthesis. When generating widely-separated views, MV-AR can utilize all its preceding views to extract effective reference information. Subsequently, we propose a unified model that accommodates various prompts via architecture designing and training strategies. To address multiple conditions, we introduce condition injection modules for text, camera pose, image, and shape. To manage multi-modal conditions simultaneously, a progressive training strategy is employed. This strategy initially adopts the text-to-multi-view (t2mv) model as a baseline to enhance the development of a comprehensive X-to-multi-view (X2mv) model through the randomly dropping and combining conditions. Finally, to alleviate the overfitting problem caused by limited high-quality data, we propose the ``Shuffle View" data augmentation technique, thus significantly expanding the training data by several magnitudes. Experiments demonstrate the performance and versatility of our MV-AR, which consistently generates consistent multi-view images across a range of conditions and performs on par with leading diffusion-based multi-view image generation models. The code and models are released at https://github.com/MILab-PKU/MVAR.

Xinliang Zhang, Lei Zhu, Hangzhou He et al.

Multimodal Large Language Models (MLLMs) have demonstrated substantial value in unified text-image understanding and reasoning, primarily by converting images into sequences of patch-level tokens that align with their architectural paradigm. However, patch-level tokenization leads to a quadratic growth in image tokens, burdening MLLMs' understanding and reasoning with enormous computation and memory. Additionally, the traditional patch-wise scanning tokenization workflow misaligns with the human vision cognition system, further leading to hallucination and computational redundancy. To address this issue, we propose an object-level token merging strategy for Adaptive Token compression, revealing the consistency with human vision system. The experiments are conducted on multiple comprehensive benchmarks, which show that our approach averagely, utilizes only 10% tokens while achieving almost 96% of the vanilla model's performance. More extensive experimental results in comparison with relevant works demonstrate the superiority of our method in balancing compression ratio and performance. Our code will be available.

JiaKui Hu, Zhengjian Yao, Lujia Jin et al.

This study introduces a Masked Degradation Classification Pre-Training method (MaskDCPT), designed to facilitate the classification of degradation types in input images, leading to comprehensive image restoration pre-training. Unlike conventional pre-training methods, MaskDCPT uses the degradation type of the image as an extremely weak supervision, while simultaneously leveraging the image reconstruction to enhance performance and robustness. MaskDCPT includes an encoder and two decoders: the encoder extracts features from the masked low-quality input image. The classification decoder uses these features to identify the degradation type, whereas the reconstruction decoder aims to reconstruct a corresponding high-quality image. This design allows the pre-training to benefit from both masked image modeling and contrastive learning, resulting in a generalized representation suited for restoration tasks. Benefit from the straightforward yet potent MaskDCPT, the pre-trained encoder can be used to address universal image restoration and achieve outstanding performance. Implementing MaskDCPT significantly improves performance for both convolution neural networks (CNNs) and Transformers, with a minimum increase in PSNR of 3.77 dB in the 5D all-in-one restoration task and a 34.8% reduction in PIQE compared to baseline in real-world degradation scenarios. It also emergences strong generalization to previously unseen degradation types and levels. In addition, we curate and release the UIR-2.5M dataset, which includes 2.5 million paired restoration samples across 19 degradation types and over 200 degradation levels, incorporating both synthetic and real-world data. The dataset, source code, and models are available at https://github.com/MILab-PKU/MaskDCPT.

Xin Li, Yeying Jin, Xin Jin et al.

This paper reviews the NTIRE 2025 Challenge on Day and Night Raindrop Removal for Dual-Focused Images. This challenge received a wide range of impressive solutions, which are developed and evaluated using our collected real-world Raindrop Clarity dataset. Unlike existing deraining datasets, our Raindrop Clarity dataset is more diverse and challenging in degradation types and contents, which includes day raindrop-focused, day background-focused, night raindrop-focused, and night background-focused degradations. This dataset is divided into three subsets for competition: 14,139 images for training, 240 images for validation, and 731 images for testing. The primary objective of this challenge is to establish a new and powerful benchmark for the task of removing raindrops under varying lighting and focus conditions. There are a total of 361 participants in the competition, and 32 teams submitting valid solutions and fact sheets for the final testing phase. These submissions achieved state-of-the-art (SOTA) performance on the Raindrop Clarity dataset. The project can be found at https://lixinustc.github.io/CVPR-NTIRE2025-RainDrop-Competition.github.io/.

Zhenxin Lei, Man Yao, Jiakui Hu et al.

Spiking Neural Networks (SNNs) have a low-power advantage but perform poorly in image segmentation tasks. The reason is that directly converting neural networks with complex architectural designs for segmentation tasks into spiking versions leads to performance degradation and non-convergence. To address this challenge, we first identify the modules in the architecture design that lead to the severe reduction in spike firing, make targeted improvements, and propose Spike2Former architecture. Second, we propose normalized integer spiking neurons to solve the training stability problem of SNNs with complex architectures. We set a new state-of-the-art for SNNs in various semantic segmentation datasets, with a significant improvement of +12.7% mIoU and 5.0 efficiency on ADE20K, +14.3% mIoU and 5.2 efficiency on VOC2012, and +9.1% mIoU and 6.6 efficiency on CityScapes.

JiaKui Hu, Zhengjian Yao, Lujia Jin et al. · pku

Translation equivariance is a fundamental inductive bias in image restoration, ensuring that translated inputs produce translated outputs. Attention mechanisms in modern restoration transformers undermine this property, adversely impacting both training convergence and generalization. To alleviate this issue, we propose two key strategies for incorporating translation equivariance: slide indexing and component stacking. Slide indexing maintains operator responses at fixed positions, with sliding window attention being a notable example, while component stacking enables the arrangement of translation-equivariant operators in parallel or sequentially, thereby building complex architectures while preserving translation equivariance. However, these strategies still create a dilemma in model design between the high computational cost of self-attention and the fixed receptive field associated with sliding window attention. To address this, we develop an adaptive sliding indexing mechanism to efficiently select key-value pairs for each query, which are then concatenated in parallel with globally aggregated key-value pairs. The designed network, called the Translation Equivariance Adaptive Transformer (TEAFormer), is assessed across a variety of image restoration tasks. The results highlight its superiority in terms of effectiveness, training convergence, and generalization.

Liying Yang, Jialun Liu, Jiakui Hu et al.

Generating high-quality 4D objects with spatial-temporal consistency is still formidable. Existing diffusion-based methods often struggle with spatial-temporal inconsistency, as they fail to leverage outputs from all previous timesteps to guide the generation at the current timestep. Therefore, we propose a Spatial-Temporal State Propagation AutoRegressive Model (4DSTAR), which generates 4D objects maintaining temporal-spatial consistency. 4DSTAR formulates the generation problem as the prediction of tokens that represent the 4D object. It consists of two key components: (1) The dynamic spatial-temporal state propagation autoregressive model (STAR) is proposed, which achieves spatial-temporal consistent generation. Unlike standard autoregressive models, STAR divides prediction tokens into groups based on timesteps. It models long-term dependencies by propagating spatial-temporal states from previous groups and utilizes these dependencies to guide generation at the next timestep. To this end, a spatial-temporal container is proposed, which dynamically updating the effective spatial-temporal state features from all historical groups, then updated features serve as conditional features to guide the prediction of the next token group. (2) The 4D VQ-VAE is proposed, which implicitly encodes the 4D structure into discrete space and decodes the discrete tokens predicted by STAR into temporally coherent dynamic 3D Gaussians. Experiments demonstrate that 4DSTAR generates spatial-temporal consistent 4D objects, and achieves performance competitive with diffusion models.

Hangzhou He, Lei Zhu, Kaiwen Li et al. · pku

Concept Bottleneck Models (CBMs) provide inherent interpretability by first predicting a set of human-understandable concepts and then mapping them to labels through a simple classifier. While users can intervene in the concept space to improve predictions, traditional CBMs typically employ a fixed linear classifier over concept scores, which restricts interventions to manual value adjustments and prevents the incorporation of new concepts or domain knowledge at test time. These limitations are particularly severe in unsupervised CBMs, where concept activations are often noisy and densely activated, making user interventions ineffective. We introduce Chat-CBM, which replaces score-based classifiers with a language-based classifier that reasons directly over concept semantics. By grounding prediction in the semantic space of concepts, Chat-CBM preserves the interpretability of CBMs while enabling richer and more intuitive interventions, such as concept correction, addition or removal of concepts, incorporation of external knowledge, and high-level reasoning guidance. Leveraging the language understanding and few-shot capabilities of frozen large language models, Chat-CBM extends the intervention interface of CBMs beyond numerical editing and remains effective even in unsupervised settings. Experiments on nine datasets demonstrate that Chat-CBM achieves higher predictive performance and substantially improves user interactivity while maintaining the concept-based interpretability of CBMs.