The evolution of Large Vision-Language Models (LVLMs) has progressed from single-image understanding to multi-image reasoning. Despite this advancement, our findings indicate that LVLMs struggle to robustly utilize information across multiple images, with predictions significantly affected by the alteration of image positions. To further explore this issue, we introduce Position-wise Question Answering (PQA), a meticulously designed task to quantify reasoning capabilities at each position. Our analysis reveals a pronounced position bias in LVLMs: open-source models excel in reasoning with images positioned later but underperform with those in the middle or at the beginning, while proprietary models like GPT-4o show improved comprehension for images at the beginning and end but struggle with those in the middle. Motivated by these insights, we propose SoFt Attention (SoFA), a simple, training-free approach that mitigates this bias by employing linear interpolation between inter-image causal attention and bidirectional counterparts. Experimental results demonstrate that SoFA effectively reduces position bias and significantly enhances the reasoning performance of existing LVLMs.
论文检索
输入标题、作者或关键词，从 9,654 篇学术成果中精准定位
Model merging combines expert models---each finetuned from a shared foundation model on diverse tasks and domains---into a single, more capable base model. However, existing model merging approaches assume all experts to be available simultaneously. In reality, new tasks and domains emerge continuously, prompting the need for a dynamic process of integrating these experts over time, which we call temporal model merging. The temporal dimension introduces unique challenges not addressed in prior work: At each task, should expert training start from merged previous experts or the original base model? Should all models be merged at every time step? Which merging techniques are best suited for temporal merging? Should different strategies be used for the training initialization and deployment phases? To tackle these questions, we propose a unified framework called TIME---Temporal Integration of Model Expertise---that defines temporal model merging across three axes: (1) Initialization Phase, (2) Deployment Phase, and (3) Merging Technique. Utilizing TIME, we study temporal model merging across model sizes, tasks, and compute budgets on the large-scale FoMo-in-Flux benchmark for continual multimodal pretraining. Systematic experiments across TIME and FoMo-in-Flux allow us to arrive at several crucial key insights for temporal model merging to better understand current limits and best practices for successful model merging across time.
Existing multimodal large language models (MLLMs) face significant challenges in Referring Video Object Segmentation(RVOS). We identify three critical challenges: (C1) insufficient quantitative representation of textual numerical data, (C2) repetitive and degraded response templates for spatiotemporal referencing, and (C3) loss of visual information in video sampling queries lacking textual guidance. To address these, we propose a novel framework, Dynamic Time Object Sensing (DTOS), specifically designed for RVOS. To tackle (C1) and (C2), we introduce specialized tokens to construct multi-answer response templates, enabling regression of event boundaries and target localization. This approach improves the accuracy of numerical regression while mitigating the issue of repetitive degradation. To address (C3), we propose a Text-guided Clip Sampler (TCS) that selects video clips aligned with user instructions, preventing visual information loss and ensuring consistent temporal resolution. TCS is also applicable to Moment Retrieval tasks, with enhanced multimodal input sequences preserving spatial details and maximizing temporal resolution. DTOS demonstrates exceptional capability in flexibly localizing multiple spatiotemporal targets based on user-provided textual instructions. Extensive experiments validate the effectiveness of our approach, with DTOS achieving state-of-the-art performance in J&F scores: an improvement of +4.36 on MeViS, +4.48 on Ref-DAVIS17, and +3.02 on Ref-YT-VOS. Additionally, our TCS demonstrates exceptional performance in Moment Retrieval. The code is available at https://github.com/Maulog/OPEN-DTOS-LMM
IceDiff: High Resolution and High-Quality Arctic Sea Ice Forecasting with Generative Diffusion Prior
PDF ↗Variation of Arctic sea ice has significant impacts on polar ecosystems, transporting routes, coastal communities, and global climate. Tracing the change of sea ice at a finer scale is paramount for both operational applications and scientific studies. Recent pan-Arctic sea ice forecasting methods that leverage advances in artificial intelligence have made promising progress over numerical models. However, forecasting sea ice at higher resolutions is still under-explored. To bridge the gap, we propose a two-module cooperative deep learning framework, IceDiff, to forecast sea ice concentration at finer scales. IceDiff first leverages a vision transformer to generate coarse yet superior forecasting results over previous methods at a regular 25km grid. This high-quality sea ice forecasting can be utilized as reliable guidance for the next module. Subsequently, an unconditional diffusion model pre-trained on low-resolution sea ice concentration maps is utilized for sampling down-scaled sea ice forecasting via a zero-shot guided sampling strategy and a patch-based method. For the first time, IceDiff demonstrates sea ice forecasting with a 6.25km resolution. IceDiff extends the boundary of existing sea ice forecasting models and more importantly, its capability to generate high-resolution sea ice concentration data is vital for pragmatic usages and research.
The distortion-perception (DP) tradeoff reveals a fundamental conflict between distortion metrics (e.g., MSE and PSNR) and perceptual quality. Recent research has increasingly concentrated on evaluating denoising algorithms within the DP framework. However, existing algorithms either prioritize perceptual quality by sacrificing acceptable distortion, or focus on minimizing MSE for faithful restoration. When the goal shifts or noisy measurements vary, adapting to different points on the DP plane needs retraining or even re-designing the model. Inspired by recent advances in solving inverse problems using score-based generative models, we explore the potential of flexibly and optimally traversing DP tradeoffs using a single pre-trained score-based model. Specifically, we introduce a variance-scaled reverse diffusion process and theoretically characterize the marginal distribution. We then prove that the proposed sample process is an optimal solution to the DP tradeoff for conditional Gaussian distribution. Experimental results on two-dimensional and image datasets illustrate that a single score network can effectively and flexibly traverse the DP tradeoff for general denoising problems.
EmotiveTalk: Expressive Talking Head Generation through Audio Information Decoupling and Emotional Video Diffusion
PDF ↗Diffusion models have revolutionized the field of talking head generation, yet still face challenges in expressiveness, controllability, and stability in long-time generation. In this research, we propose an EmotiveTalk framework to address these issues. Firstly, to realize better control over the generation of lip movement and facial expression, a Vision-guided Audio Information Decoupling (V-AID) approach is designed to generate audio-based decoupled representations aligned with lip movements and expression. Specifically, to achieve alignment between audio and facial expression representation spaces, we present a Diffusion-based Co-speech Temporal Expansion (Di-CTE) module within V-AID to generate expression-related representations under multi-source emotion condition constraints. Then we propose a well-designed Emotional Talking Head Diffusion (ETHD) backbone to efficiently generate highly expressive talking head videos, which contains an Expression Decoupling Injection (EDI) module to automatically decouple the expressions from reference portraits while integrating the target expression information, achieving more expressive generation performance. Experimental results show that EmotiveTalk can generate expressive talking head videos, ensuring the promised controllability of emotions and metric stability during long-time generation, yielding state-of-the-art performance compared to existing methods. The main page of our paper can be found in https://emotivetalk.github.io/.
Black-box Domain Adaptation (BDA) utilizes a black-box predictor of the source domain to label target domain data, addressing privacy concerns in Unsupervised Domain Adaptation (UDA). However, BDA assumes identical label sets across domains, which is unrealistic. To overcome this limitation, we propose a study on BDA with unknown classes in the target domain. It uses a black-box predictor to label target data and identify "unknown" categories, without requiring access to source domain data or predictor parameters, thus addressing both data privacy and category shift issues in traditional UDA. Existing methods face two main challenges: (i) Noisy pseudo-labels in knowledge distillation (KD) accumulate prediction errors, and (ii) relying on a preset threshold fails to adapt to varying category shifts. To address these, we propose ADU, a framework that allows the target domain to autonomously learn pseudo-labels guided by quality and use an adaptive threshold to identify "unknown" categories. Specifically, ADU consists of Selective Amplification Knowledge Distillation (SAKD) and Entopy-Driven Label Differentiation (EDLD). SAKD improves KD by focusing on high-quality pseudo-labels, mitigating the impact of noisy labels. EDLD categorizes pseudo-labels by quality and applies tailored training strategies to distinguish "unknown" categories, improving detection accuracy and adaptability. Extensive experiments show that ADU achieves state-of-the-art results, outperforming the best existing method by 3.1% on VisDA in the OPBDA scenario.
We present a technique and benchmark dataset for estimating the relative 3D orientation between a pair of Internet images captured in an extreme setting, where the images have limited or non-overlapping field of views. Prior work targeting extreme rotation estimation assume constrained 3D environments and emulate perspective images by cropping regions from panoramic views. However, real images captured in the wild are highly diverse, exhibiting variation in both appearance and camera intrinsics. In this work, we propose a Transformer-based method for estimating relative rotations in extreme real-world settings, and contribute the ExtremeLandmarkPairs dataset, assembled from scene-level Internet photo collections. Our evaluation demonstrates that our approach succeeds in estimating the relative rotations in a wide variety of extreme-view Internet image pairs, outperforming various baselines, including dedicated rotation estimation techniques and contemporary 3D reconstruction methods.
MAtCha Gaussians: Atlas of Charts for High-Quality Geometry and Photorealism From Sparse Views
PDF ↗We present a novel appearance model that simultaneously realizes explicit high-quality 3D surface mesh recovery and photorealistic novel view synthesis from sparse view samples. Our key idea is to model the underlying scene geometry Mesh as an Atlas of Charts which we render with 2D Gaussian surfels (MAtCha Gaussians). MAtCha distills high-frequency scene surface details from an off-the-shelf monocular depth estimator and refines it through Gaussian surfel rendering. The Gaussian surfels are attached to the charts on the fly, satisfying photorealism of neural volumetric rendering and crisp geometry of a mesh model, i.e., two seemingly contradicting goals in a single model. At the core of MAtCha lies a novel neural deformation model and a structure loss that preserve the fine surface details distilled from learned monocular depths while addressing their fundamental scale ambiguities. Results of extensive experimental validation demonstrate MAtCha's state-of-the-art quality of surface reconstruction and photorealism on-par with top contenders but with dramatic reduction in the number of input views and computational time. We believe MAtCha will serve as a foundational tool for any visual application in vision, graphics, and robotics that require explicit geometry in addition to photorealism.
Zero-shot 4D segmentation and recognition of arbitrary objects in Lidar is crucial for embodied navigation, with applications ranging from streaming perception to semantic mapping and localization. However, the primary challenge in advancing research and developing generalized, versatile methods for spatio-temporal scene understanding in Lidar lies in the scarcity of datasets that provide the necessary diversity and scale of annotations. To overcome these challenges, we propose SAL-4D (Segment Anything in Lidar--4D), a method that utilizes multi-modal robotic sensor setups as a bridge to distill recent developments in Video Object Segmentation (VOS) in conjunction with off-the-shelf Vision-Language foundation models to Lidar. We utilize VOS models to pseudo-label tracklets in short video sequences, annotate these tracklets with sequence-level CLIP tokens, and lift them to the 4D Lidar space using calibrated multi-modal sensory setups to distill them to our SAL-4D model. Due to temporal consistent predictions, we outperform prior art in 3D Zero-Shot Lidar Panoptic Segmentation (LPS) over 5 PQ, and unlock Zero-Shot 4D-LPS.
3D Question Answering (3D QA) requires the model to comprehensively understand its situated 3D scene described by the text, then reason about its surrounding environment and answer a question under that situation. However, existing methods usually rely on global scene perception from pure 3D point clouds and overlook the importance of rich local texture details from multi-view images. Moreover, due to the inherent noise in camera poses and complex occlusions, there exists significant feature degradation and reduced feature robustness problems when aligning 3D point cloud with multi-view images. In this paper, we propose a Dual-vision Scene Perception Network (DSPNet), to comprehensively integrate multi-view and point cloud features to improve robustness in 3D QA. Our Text-guided Multi-view Fusion (TGMF) module prioritizes image views that closely match the semantic content of the text. To adaptively fuse back-projected multi-view images with point cloud features, we design the Adaptive Dual-vision Perception (ADVP) module, enhancing 3D scene comprehension. Additionally, our Multimodal Context-guided Reasoning (MCGR) module facilitates robust reasoning by integrating contextual information across visual and linguistic modalities. Experimental results on SQA3D and ScanQA datasets demonstrate the superiority of our DSPNet. Codes will be available at https://github.com/LZ-CH/DSPNet.
Most current video MLLMs rely on uniform frame sampling and image-level encoders, resulting in inefficient data processing and limited motion awareness. To address these challenges, we introduce EMA, an Efficient Motion-Aware video MLLM that utilizes compressed video structures as inputs. We propose a motion-aware GOP (Group of Pictures) encoder that fuses spatial and motion information within a GOP unit in the compressed video stream, generating compact, informative visual tokens. By integrating fewer but denser RGB frames with more but sparser motion vectors in this native slow-fast input architecture, our approach reduces redundancy and enhances motion representation. Additionally, we introduce MotionBench, a benchmark for evaluating motion understanding across four motion types: linear, curved, rotational, and contact-based. Experimental results show that EMA achieves state-of-the-art performance on both MotionBench and popular video question answering benchmarks, while reducing inference costs. Moreover, EMA demonstrates strong scalability, as evidenced by its competitive performance on long video understanding benchmarks.
Multi-view clustering represents one of the most established paradigms within the field of unsupervised learning and has witnessed a surge in popularity in recent years. View-pair form contrastive learning allows for consistently representing multiple views by maximizing mutual information between each two views. This approach permits multi-view clustering to discern consistent latent representations across multiple views. However, two significant issues emerge when this approach is considered. i)it is challenging to ascertain which two views are most appropriate for contrastive learning when there are more than three views, particularly without prior knowledge. ii) when all views are included in contrastive learning, multi-view clustering performance is compromised by poor quality views. To tackle these issues, we present a novel Efficient Dual Selection Mechanism for deep Multi-View Clustering framework, termed EASEMVC. Specifically, EASEMVC first constructs a view graph based on the OT distance between the bipartite graph of each view. It then designs a view selection module to realize an efficient view-level selection process through the view topology relations in the view graph structure. Additionally, a cross-view sample graph structure is constructed at the sample level, with the sample topological relations in the cross-view sample graph structure being employed to generate reliable sample learning weights. Based on the view pairs and sample weights selected by the aforementioned method, we employ contrastive learning within the theoretical framework of information theory to obtain consistent representations between views. Extensive experiments on six benchmark datasets have demonstrated that the proposed EASEMVC outperforms the state-of-the-art methods.
Hyperspectral pansharpening refers to fusing a panchromatic image (PAN) and a low-resolution hyperspectral image (LR-HSI) to obtain a high-resolution hyperspectral image (HR-HSI). Recently, guiding pre-trained diffusion models (DMs) has demonstrated significant potential in this area, leveraging their powerful representational abilities while avoiding complex training processes. However, these DMs are often trained on RGB images, not well-suited for pansharpening tasks, limited in adapting to the hyperspectral images. In this work, we propose a novel guided diffusion scheme with zero-shot guidance and neural spatial-spectral decomposition (NSSD) to iteratively generate the RGB detail image and map the RGB detail image to target HR-HSI. Specifically, zero-shot guidance employs an auxiliary neural network that trained only with a PAN and LR-HSI to guide pre-trained DMs in generating the RGB detail image, informed by specific prior knowledge. Then, NSSD establishes a spectral mapping from the generated RGB detail image to the final HR-HSI. Extensive experiments are conducted on Pavia, Washington DC, Chukusei, and FR1 datasets to demonstrate that the proposed method significantly enhances the performance of DMs for hyperspectral pansharpening tasks, outperforming existing methods across multiple metrics and achieving improvements in visualization results. The code is available at \href https://github.com/Jin-liangXiao/DM-zs https://github.com/Jin-liangXiao/DM-zs .
Comprehensive Relighting: Generalizable and Consistent Monocular Human Relighting and Harmonization
PDF ↗This paper introduces Comprehensive Relighting, the first all-in-one approach that can both control and harmonize the lighting from an image or video of humans with arbitrary body parts from any scene. Building such a generalizable model is extremely challenging due to the lack of dataset, restricting existing image-based relighting models to a specific scenario (e.g., face or static human). To address this challenge, we repurpose a pre-trained diffusion model as a general image prior and jointly model the human relighting and background harmonization in the coarse-to-fine framework. To further enhance the temporal coherence of the relighting, we introduce an unsupervised temporal lighting model that learns the lighting cycle consistency from many real-world videos without any ground truth. In inference time, our temporal lighting module is combined with the diffusion models through the spatio-temporal feature blending algorithms without extra training; and we apply a new guided refinement as a post-processing to preserve the high-frequency details from the input image. In the experiments, Comprehensive Relighting shows a strong generalizability and lighting temporal coherence, outperforming existing image-based human relighting and harmonization methods.
Feature4X: Bridging Any Monocular Video to 4D Agentic AI with Versatile Gaussian Feature Fields
PDF ↗Recent advancements in 2D and multimodal models have achieved remarkable success by leveraging large-scale training on extensive datasets. However, extending these achievements to enable free-form interactions and high-level semantic operations with complex 3D/4D scenes remains challenging. This difficulty stems from the limited availability of large-scale, annotated 3D/4D or multi-view datasets, which are crucial for generalizable vision and language tasks such as open-vocabulary and prompt-based segmentation, language-guided editing, and visual question answering (VQA). In this paper, we introduce Feature4X, a universal framework designed to extend any functionality from 2D vision foundation model into the 4D realm, using only monocular video input, which is widely available from user-generated content. The "X" in Feature4X represents its versatility, enabling any task through adaptable, model-conditioned 4D feature field distillation. At the core of our framework is a dynamic optimization strategy that unifies multiple model capabilities into a single representation. Additionally, to the best of our knowledge, Feature4X is the first method to distill and lift the features of video foundation models (e.g. SAM2, InternVideo2) into an explicit 4D feature field using Gaussian Splatting. Our experiments showcase novel view segment anything, geometric and appearance scene editing, and free-form VQA across all time steps, empowered by LLMs in feedback loops. These advancements broaden the scope of agentic AI applications by providing a foundation for scalable, contextually and spatiotemporally aware systems capable of immersive dynamic 4D scene interaction.
Recent synthetic 3D human datasets for the face, body, and hands have pushed the limits on photorealism. Face recognition and body pose estimation have achieved state-of-the-art performance using synthetic training data alone, but for the hand, there is still a large synthetic-to-real gap. This paper presents the first systematic study of the synthetic-to-real gap of 3D hand pose estimation. We analyze the gap and identify key components such as the forearm, image frequency statistics, hand pose, and object occlusions. To facilitate our analysis, we propose a data synthesis pipeline to synthesize high-quality data. We demonstrate that synthetic hand data can achieve the same level of accuracy as real data when integrating our identified components, paving the path to use synthetic data alone for hand pose estimation. Code and data are available at: https://github.com/delaprada/HandSynthesis.git.
Recent efforts recognize the power of scale in 3D learning (e.g. PTv3) and attention mechanisms (e.g. FlashAttention).However, current point cloud backbones fail to holistically unify geometric locality, attention mechanisms, and GPU architectures in one view.In this paper, we introduce Flash3D Transformer, which aligns geometric locality and GPU tiling through a principled locality mechanism based on Perfect Spatial Hashing (PSH).The common alignment with GPU tiling naturally fuses our PSH locality mechanism with FlashAttention at negligible extra cost.This mechanism affords flexible design choices throughout the backbone that result in superior downstream task results.Flash3D outperforms state-of-the-art PTv3 results on benchmark datasets, delivering a 2.25x speed increase and 2.4x memory efficiency boost. This efficiency enables scaling to wider attention scopes and larger models without additional overhead.Such scaling allows Flash3D to achieve even higher task accuracies than PTv3 under the same compute budget.
Weakly-supervised methods for video anomaly detection (VAD) are conventionally based merely on RGB spatio-temporal features, which continues to limit their reliability in real-world scenarios. This is due to the fact that RGB-features are not sufficiently distinctive in setting apart categories such as shoplifting from visually similar events. Therefore, towards robust complex real-world VAD, it is essential to augment RGB spatio-temporal features by additional modalities. Motivated by this, we introduce the Poly-modal Induced framework for VAD: PI-VAD (or \pi-VAD), a novel approach that augments RGB representations by five additional modalities. Specifically, the modalities include sensitivity to fine-grained motion (Pose), three dimensional scene and entity representation (Depth), surrounding objects (Panoptic masks), global motion (optical flow), as well as language cues (VLM). Each modality represents an axis of a polygon, streamlined to add salient cues to RGB. \pi-VAD includes two plug-in modules, namely Pseudo-modality Generation module and Cross Modal Induction module, which generate modality-specific prototypical representation and, thereby, induce multi-modal information into RGB cues. These modules operate by performing anomaly-aware auxiliary tasks and necessitate five modality backbones -- only during training. Notably, \pi-VAD achieves state-of-the-art accuracy on three prominent VAD datasets encompassing real-world scenarios, without requiring the computational overhead of five modality backbones at inference.
O-TPT: Orthogonality Constraints for Calibrating Test-time Prompt Tuning in Vision-Language Models
PDF ↗Test-time prompt tuning for vision-language models (VLMs) is getting attention because of their ability to learn with unlabeled data without fine-tuning. Although test-time prompt tuning methods for VLMs can boost accuracy, the resulting models tend to demonstrate poor calibration, which casts doubts on the reliability and trustworthiness of these models. Notably, more attention needs to be devoted to calibrating the test-time prompt tuning in vision-language models. To this end, we propose a new approach, called O-TPT that introduces orthogonality constraints on the textual features corresponding to the learnable prompts for calibrating test-time prompt tuning in VLMs. Towards introducing orthogonality constraints, we make the following contributions. First, we uncover new insights behind the suboptimal calibration performance of existing methods relying on textual feature dispersion. Second, we show that imposing a simple orthogonalization of textual features is a more effective approach towards obtaining textual dispersion. We conduct extensive experiments on various datasets with different backbones and baselines. The results indicate that our method consistently outperforms the prior state of the art in significantly reducing the overall average calibration error. Also, our method surpasses the zero-shot calibration performance on fine-grained classification tasks. Our code is available at https://github.com/ashshaksharifdeen/O-TPT.