Multimodal large language models (MLLMs) are broadly empowering various fields. Despite their advancements, the explainability of MLLMs remains less explored, hindering deeper understanding, model credibility, and effective visualization. Unlike conventional vision models (e.g., CNNs, ViTs, CLIP) that produce a single output, MLLMs generate sequences of tokens progressively, where each generated token depends on the previous context. Therefore, earlier context tokens can introduce redundant activations that interfere with the explanation of later tokens beyond their original information. Existing studies often overlook this issue, but our observations reveal that these redundant correlations can significantly hurt the reliability of explanations. To address this, we propose an estimated causal inference method to mitigate the interference of context to achieve high-quality MLLM explanation, with a novel rank Gaussian filter to further reduce activation noises. We term this method Token Activation Map (TAM) to highlight the consideration of interactions between tokens. TAM also indicates that it excels at explaining multiple tokens of MLLM, which is different from the Class Activation Map (CAM) for a single prediction. Our TAM method significantly outperforms existing SoTA methods, showcasing high-quality visualization results that can be utilized for various scenarios, such as object localization, failure case analysis, video visualization, MLLMs visual comparison, and model understanding (e.g., color, shape, action, location, visual reasoning, multi-turn conversation, etc). The code is available at github.com/xmed-lab/TAM.
论文检索
输入标题、作者或关键词，从 2,701 篇学术成果中精准定位
Dataset distillation (DD) condenses large datasets into compact yet informative substitutes, preserving performance comparable to the original dataset while reducing storage, transmission costs, and computational consumption. However, previous DD methods mainly focus on distilling information from images, often overlooking the semantic information inherent in the data. The disregard for context hinders the model's generalization ability, particularly in tasks involving complex datasets, which may result in illogical outputs or the omission of critical objects. In this study, we integrate vision-language methods into DD by introducing text prototypes to distill language information and collaboratively synthesize data with image prototypes, thereby enhancing dataset distillation performance. Notably, the text prototypes utilized in this study are derived from descriptive text information generated by an open-source vision-language model. This framework demonstrates broad applicability across datasets without pre-existing text descriptions, expanding the potential of dataset distillation beyond traditional image-based approaches. Compared to other methods, the proposed approach generates logically coherent images containing target objects, achieving state-of-the-art validation performance and demonstrating robust generalization. Source code and generated data are available in https://github.com/zou-yawen/Dataset-Distillation-via-Vision-Language-Category-Prototype/.
CityGS-X: A Scalable Architecture for Efficient and Geometrically Accurate Large-Scale Scene Reconstruction
PDF ↗Despite its significant achievements in large-scale scene reconstruction, 3D Gaussian Splatting still faces substantial challenges, including slow processing, high computational costs, and limited geometric accuracy. These core issues arise from its inherently unstructured design and the absence of efficient parallelization. To overcome these challenges simultaneously, we introduce CityGS-X, a scalable architecture built on a novel parallelized hybrid hierarchical 3D representation (PH2-3D). As an early attempt, CityGS-X abandons the cumbersome merge-and-partition process and instead adopts a newly designed batch-level multi-task rendering process. This architecture enables efficient multi-GPU rendering through dynamic Level-of-Detail voxel allocations, significantly improving scalability and performance. Through extensive experiments, CityGS-X consistently outperforms existing methods in terms of faster training times, larger rendering capacities, and more accurate geometric details in large-scale scenes. To further enhance both overall quality and geometric accuracy, CityGS-X presents a progressive RGB-Depth-Normal training strategy. This approach enhances 3D consistency by jointly optimizing appearance and geometry representation through multi-view constraints and off-the-shelf depth priors within batch-level training. Notably, CityGS-X can train and render a scene with more than 5000 images in just five hours using only four 4090 GPUs, a task that would make other alternative methods encounter out-of-memory issues and fail completely. This implies that CityGS-X is far beyond the capacity of other existing methods.
Sufficient cross-task interaction is crucial for success in multi-task dense prediction. However, sufficient interaction often results in high computational complexity, forcing existing methods to face the trade-off between interaction completeness and computational efficiency. To address this limitation, this work proposes a Bidirectional Interaction Mamba (BIM), which incorporates novel scan mechanisms to adapt the Mamba modeling approach for multi-task dense prediction. On the one hand, we introduce a novel Bidirectional Interaction Scan (BI-Scan) mechanism, which constructs task-specific representations as bidirectional sequences during interaction. By integrating task-first and position-first scan modes within a unified linear complexity architecture, BI-Scan efficiently preserves critical cross-task information. On the other hand, we employ a Multi-Scale Scan (MS-Scan) mechanism to achieve multi-granularity scene modeling. This design not only meets the diverse granularity requirements of various tasks but also enhances nuanced cross-task feature interactions. Extensive experiments on two challenging benchmarks, i.e., NYUD-V2 and PASCAL-Context, show the superiority of our BIM vs its state-of-the-art competitors.
Editing images via instruction provides a natural way to generate interactive content, but it is a big challenge due to the higher requirement of scene understanding and generation. Prior work utilizes a chain of large language models, object segmentation models, and editing models for this task. However, the understanding models provide only single-modality ability, restricting the editing quality. We aim to bridge understanding and generation via a new multi-modality model that provides the intelligent abilities to instruction-based image editing models for more complex cases. To achieve this goal, we individually separate the instruction editing task with the multi-modality chain of thought prompts, i.e., Chain-of-Thought (CoT) planning, editing region reasoning, and editing. For Chain-of-Thought planning, the large language model could reason the appropriate sub-prompts considering the instruction provided and the ability of the editing network. For editing region reasoning, we train an instruction-based editing region generation network with a multi-modal large language model. Finally, a hint-guided instruction-based editing network is proposed for editing image generations based on the sizeable text-to-image diffusion model to accept the hints for generation. Extensive experiments demonstrate that our method has competitive editing abilities on complex real-world images. The source code will be publicly available.
Current visual foundation models (VFMs) face a fundamental limitation in transferring knowledge from vision language models (VLMs): while VLMs excel at modeling cross-modal interactions through unified representation spaces, existing VFMs predominantly adopt result-oriented paradigms that neglect the underlying interaction processes. This representational discrepancy leads to suboptimal knowledge transfer and limited generalization capabilities across vision tasks.We propose Learning from Interactions, a cognitive-inspired framework that bridges this gap by explicitly modeling interactions during visual understanding. Our key insight is that preserving the interaction dynamics captured by VLMs -- rather than just their final representations -- enables more effective knowledge transfer to downstream VFMs. The technical core involves two innovations: (1) Interaction Queries that maintain persistent relationships across network layers, and (2) interaction-based supervision derived from pre-trained VLMs' cross-modal attention patterns.Comprehensive experiments demonstrate consistent improvements across multiple benchmarks: achieving ~3.3% and +1.6 mAP/+2.4 AP^ mask absolute gains on TinyImageNet classification and COCO detection/segmentation respectively, with minimal parameter overhead and faster convergence (7xspeedup). The framework particularly excels in cross-domain scenarios, delivering ~2.4% and ~9.3% zero-shot improvements on PACS and VLCS. Human evaluations confirm our approach's cognitive alignment, outperforming result-oriented methods by 2.7xin semantic consistency metrics.
FedWSQ: Efficient Federated Learning with Weight Standardization and Distribution-Aware Non-Uniform Quantization
PDF ↗Federated learning (FL) often suffers from performance degradation due to key challenges such as data heterogeneity and communication constraints. To address these limitations, we present a novel FL framework called FedWSQ, which integrates weight standardization (WS) and the proposed distribution-aware non-uniform quantization (DANUQ). WS enhances FL performance by filtering out biased components in local updates during training, thereby improving the robustness of the model against data heterogeneity and unstable client participation. In addition, DANUQ minimizes quantization errors by leveraging the statistical properties of local model updates. As a result, FedWSQ significantly reduces communication overhead while maintaining superior model accuracy. Extensive experiments on FL benchmark datasets demonstrate that FedWSQ consistently outperforms existing FL methods across various challenging FL settings, including extreme data heterogeneity and ultra-low-bit communication scenarios. The source code is available at our project page.
We build a comprehensive online evaluation benchmark for language-conditioned multi-step task execution on mobile interfaces. Our benchmark strives to evaluate the multi-step planning, reasoning, and visual grounding capabilities of agents, using mobile user interfaces as a concrete testbed. To build diverse, challenging tasks that reflect real-world use cases, we propose an exhaustive taxonomy that allows us to measure progress along multiple decision-making abilities including multi-step planning, visual perception, action grounding, and using memory or external knowledge. We also highlight important factors such as statefulness, safety, and evaluation complexity that are key to design tasks that can be reliably evaluated. Using this taxonomy, we design 116 tasks across 36 unique apps. Through an automatic framework, we stage and evaluate several natural baselines with different input representations and planning strategies. We show that the best-performing agent achieves 40% success on our benchmark. We further measure agents' abilities to plan, ground, and utilize world knowledge highlighting areas of improvement.
Mainstream high dynamic range imaging techniques typically rely on fusing multiple images captured with different exposure setups (shutter speed and ISO). A good balance between shutter speed and ISO is crucial for achieving high-quality HDR, as high ISO values introduce significant noise, while long shutter speeds can lead to noticeable motion blur. However, existing methods often overlook the complex interaction between shutter speed and ISO and fail to account for motion blur effects in dynamic scenes. In this work, we propose AdaptiveAE, a reinforcement learning-based method that optimizes the selection of shutter speed and ISO combinations to maximize HDR reconstruction quality in dynamic environments. AdaptiveAE integrates an image synthesis pipeline that incorporates motion blur and noise simulation into our training procedure, leveraging semantic information and exposure histograms. It can adaptively select optimal ISO and shutter speed sequences based on a user-defined exposure time budget, and find a better exposure schedule than traditional solutions. Experimental results across multiple datasets demonstrate that it achieves the state-of-the-art performance.
Graphic design visually conveys information and data by creating and combining text, images and graphics. Two-stage methods that rely primarily on layout generation lack creativity and intelligence, making graphic design still labor-intensive. Existing diffusion-based methods generate non-editable graphic design files at image level with poor legibility in visual text rendering, which prevents them from achieving satisfactory and practical automated graphic design. In this paper, we propose Instructional Graphic Designer (IGD) to swiftly generate multimodal layers with editable flexibility with only natural language instructions. IGD adopts a new paradigm that leverages parametric rendering and image asset generation. First, we develop a design platform and establish a standardized format for multi-scenario design files, thus laying the foundation for scaling up data. Second, IGD utilizes the multimodal understanding and reasoning capabilities of MLLM to accomplish attribute prediction, sequencing and layout of layers. It also employs a diffusion model to generate image content for assets. By enabling end-to-end training, IGD architecturally supports scalability and extensibility in complex graphic design tasks. The superior experimental results demonstrate that IGD offers a new solution for graphic design.
We propose a method to extend foundational monocular depth estimators (FMDEs), trained on perspective images, to fisheye images. Despite being trained on tens of millions of images, FMDEs are susceptible to the covariate shift introduced by changes in camera calibration (intrinsic, distortion) parameters, leading to erroneous depth estimates. Our method aligns the distribution of latent embeddings encoding fisheye images to those of perspective images, enabling the reuse of FMDEs for fisheye cameras without retraining or finetuning. To this end, we introduce a set of Calibration Tokens as a light-weight adaptation mechanism that modulates the latent embeddings for alignment. By exploiting the already expressive latent space of FMDEs, we posit that modulating their embeddings avoids the negative impact of artifacts and loss introduced in conventional recalibration or map projection to a canonical reference frame in the image space. Our method is self-supervised and does not require fisheye images but leverages publicly available large-scale perspective image datasets. This is done by recalibrating perspective images to fisheye images, and enforcing consistency between their estimates during training. We evaluate our approach with several FMDEs, on both indoors and outdoors, where we consistently improve over state-of-the-art methods using a single set of tokens for both. Code available at: https://github.com/JungHeeKim29/calibration-token.
Generalized few-shot point cloud segmentation (GFS-3DSeg) aims to segment objects of both base and novel classes using abundant base class samples and limited novel class samples. Existing GFS-3DSeg methods encounter bottlenecks due to the scarcity of novel class data and inter-class confusion. In this paper, we propose the LLM-Assisted Hyper-Relation Matching (LARM) framework, which leverages the wealth of prior knowledge in Large Language Models (LLM) to enrich novel category prototypes and introduces a hyper-relation matching strategy to mitigate false matches between point features and category prototypes caused by inter-class confusion. The proposed LARM enjoys several merits. First, the vast knowledge embedded in LLM can be an effective complement to vanilla category prototypes, enabling them to exhibit greater robustness. Second, the hyper-relation matching strategy harnesses the structural information implicit in the inter-class relationships, making it more robust than individual feature comparisons. Extensive experiments on two benchmarks demonstrate that LARM outperforms previous state-of-the-art methods by large margins.
Video understanding requires the extraction of rich spatio-temporal representations, achieved by transformer models through self-attention. Unfortunately, self-attention poses a computational burden. In NLP, Mamba has surfaced as an efficient alternative for transformers. However, Mamba's successes do not trivially extend to vision tasks, including those in video analysis. In this paper, we theoretically analyze the differences between self-attention and Mamba. We identify two limitations in Mamba's token processing: historical decay and element contradiction. We propose VideoMambaPro (VMP) that addresses these limitations by adding masked backward computation and elemental residual connections to a VideoMamba backbone. VideoMambaPro models surpass VideoMamba by 1.6-3.0% and 1.1-1.9% top-1 on Kinetics-400 and Something-Something V2, respectively. Even without extensive pre-training, our models present an attractive and efficient alternative to current transformer models. Moreover, our two solutions are orthogonal to recent advances in Vision Mamba models, and are likely to provide further improvements in future models.
Our objective is the automatic generation of Audio Descriptions (ADs) for edited video material, such as movies and TV series. To achieve this, we propose a two-stage framework that leverages "shots" as the fundamental units of video understanding. This includes extending temporal context to neighboring shots and incorporating film grammar devices, such as shot scales and thread structures, to guide AD generation. Our method is compatible with both open-source and proprietary Visual-Language Models (VLMs), integrating expert knowledge from add-on modules without requiring additional training of the VLMs. We achieve state-of-the-art performance among all prior training-free approaches and even surpass fine-tuned methods on several benchmarks. To evaluate the quality of predicted ADs, we introduce a new evaluation measure -- an action score -- specifically targeted to assessing this important aspect of AD. Additionally, we propose a novel evaluation protocol that treats automatic frameworks as AD generation assistants and asks them to generate multiple candidate ADs for selection.
Quantization is a key technique to reduce network size and computational complexity by representing the network parameters with a lower precision. Traditional quantization methods rely on access to original training data, which is often restricted due to privacy concerns or security challenges. Zero-shot Quantization (ZSQ) addresses this by using synthetic data generated from pre-trained models, eliminating the need for real training data.Recently, ZSQ has been extended to object detection. However, existing methods use unlabeled task-agnostic synthetic images that lack the specific information required for object detection, leading to suboptimal performance. In this paper, we propose a novel task-specific ZSQ framework for object detection networks, which consists of two main stages. First, we introduce a bounding box and category sampling strategy to synthesize a task-specific calibration set from the pre-trained network, reconstructing object locations, sizes, and category distributions without any prior knowledge. Second, we integrate task-specific training into the knowledge distillation process to restore the performance of quantized detection networks.Extensive experiments conducted on the MS-COCO and Pascal VOC datasets demonstrate the efficiency and state-of-the-art performance of our method.
The practical deployment of diffusion models is still hindered by the high memory and computational overhead. Although quantization paves a way for model compression and acceleration, existing methods face challenges in achieving low-bit quantization efficiently. In this paper, we identify imbalanced activation distributions as a primary source of quantization difficulty, and propose to adjust these distributions through weight finetuning to be more quantization-friendly. We provide both theoretical and empirical evidence supporting finetuning as a practical and reliable solution. Building on this approach, we further distinguish two critical types of quantized layers: those responsible for retaining essential temporal information and those particularly sensitive to bit-width reduction. By selectively finetuning these layers under both local and global supervision, we mitigate performance degradation while enhancing quantization efficiency. Our method demonstrates its efficacy across three high-resolution image generation tasks, obtaining state-of-the-art performance across multiple bit-width settings.
We present T2Bs, a framework for generating high-quality, animatable character head morphable models from text by combining static text-to-3D generation with video diffusion. Text-to-3D models produce detailed static geometry but lack motion synthesis, while video diffusion models generate motion with temporal and multi-view geometric inconsistencies. T2Bs bridges this gap by leveraging deformable 3D Gaussian splatting to align static 3D assets with video outputs. By constraining motion with static geometry and employing a view-dependent deformation MLP, T2Bs (i) outperforms existing 4D generation methods in accuracy and expressiveness while reducing video artifacts and view inconsistencies, and (ii) reconstructs smooth, coherent, fully registered 3D geometries designed to scale for building morphable models with diverse, realistic facial motions. This enables synthesizing expressive, animatable character heads that surpass current 4D generation techniques.
Large language models (LLM) in natural language processing (NLP) have demonstrated great potential for in-context learning (ICL) -- the ability to leverage a few set of example prompts to adapt to various tasks without having to explicitly update model weights. ICL has recently been explored for the visual domain with promising early outcomes. These approaches involve specialized training and/or additional data which complicate the process and limit its generalizability. In this work, we show that off-the-shelf Stable Diffusion models can be repurposed for visual in-context learning (V-ICL). Specifically, we formulate an in-place attention re-computation within the self-attention layers of the Stable Diffusion architecture that explicitly incorporates context between the query and example prompts. Without any additional fine-tuning, we show that this repurposed Stable Diffusion model is able to adapt to six different tasks: foreground segmentation, single object detection, semantic segmentation, keypoint detection, edge detection, and colorization. For example, the proposed approach improves the mean intersection over union (mIoU) for the foreground segmentation task on Pascal-5i dataset by 8.9% and 3.2% over recent methods such as Visual Prompting and IMProv, respectively. Additionally, we show that the proposed method is able to effectively leverage multiple prompts through ensembling to infer the task better and further improve the performance.
As a verified need, consistent editing across in-the-wild images remains a technical challenge arising from various unmanageable factors, like object poses, lighting conditions, and photography environments. Edicho steps in with a training-free solution based on diffusion models, featuring a fundamental design principle of using explicit image correspondence to direct editing. Specifically, the key components include an attention manipulation module and a carefully refined classifier-free guidance (CFG) denoising strategy, both of which take into account the pre-estimated correspondence. Such an inference-time algorithm enjoys a plug-and-play nature and is compatible to most diffusion-based editing methods, such as ControlNet and BrushNet. Extensive results demonstrate the efficacy of Edicho in consistent cross-image editing under diverse settings.
We propose DepR, a depth-guided single-view scene reconstruction framework that integrates instance-level diffusion within a compositional paradigm. Instead of reconstructing the entire scene holistically, DepR generates individual objects and subsequently composes them into a coherent 3D layout. Unlike previous methods that use depth solely for object layout estimation during inference and therefore fail to fully exploit its rich geometric information, DepR leverages depth throughout both training and inference. Specifically, we introduce depth-guided conditioning to effectively encode shape priors into diffusion models. During inference, depth further guides DDIM sampling and layout optimization, enhancing alignment between the reconstruction and the input image. Despite being trained on limited synthetic data, DepR achieves state-of-the-art performance and demonstrates strong generalization in single-view scene reconstruction, as shown through evaluations on both synthetic and real-world datasets.