AAAI 2026 Papers

Self-supervised 3D point cloud understanding is crucial for scene understanding, where Masked Autoencoders (MAE) have achieved excellent performance in point cloud representation learning. However, existing MAE-style methods fail to consider spatial-semantic variations in masking strategies, and joint learning with multi-view images often overlooks view redundancy. To address these challenges, we propose an MAE framework enhanced with reliable multi-view 2D-3D Key-part alignment and Reinforced masking, named as KR-MAE. Our approach comprises three key innovations: Reinforced Masking (RM) strategically samples visible tokens based on semantic saliency to enhance reconstruction fidelity; Reliable Multi-View Selector (RVS) dynamically refines the most informative image subset by filtering occluded or low-texture views, mitigating detrimental redundancy; Reliable-view 2D-3D Key-part Aligned Transformer (KAT) establishes semantic-aligned correspondence between salient 3D point cloud parts and reliable multi-view 2D image patches, leveraging rich texture cues from 2D images to compensate for sparse geometry in point cloud. Extensive experiments on 3D classification and segmentation benchmarks demonstrate that KR-MAE achieves state-of-the-art performance, surpassing prior multi-modal methods.

Versatile 3D tasks (e.g., generation or editing) distilling Text-to-Image (T2I) diffusion models have attracted significant research interest for not relying on extensive 3D training data. However, T2I models exhibit limitations resulting from prior view bias, which produces conflicting appearances between different views of an object. This bias causes subject-words to preferentially activate prior view features during cross-attention (CA) computation, regardless of the target view condition. To overcome this limitation, we conduct a comprehensive mathematical analysis to reveal the root cause of the prior view bias in T2I models. Moreover, we find different UNet-Layers show different effects of prior view in CA. Therefore, we propose a novel framework, TD-Attn, which addresses multi-view inconsistency via two key components: (1) the 3D-Aware Attention Guidance Module 3D-AAG constructs a view-consistent 3D attention Gaussian for subject-words to enforce spatial consistency across attention-focused regions, thereby compensating for the limited spatial information in 2D individual view CA maps; (2) the Hierarchical Attention Modulation Module (HAM) utilizes a semantic guidance tree to direct the Semantic Response Profiler (SRP) in localizing and modulating CA layers that are highly responsive to view conditions, where the enhanced CA maps further support the construction of more consistent 3D attention Gaussians. Notably, HAM facilitates semantic-specific interventions, enabling controllable and precise 3D editing. Extensive experiments firmly establish that TD-Attn has the potential to serve as a transformative, universal plugin, significantly enhancing multi-view consistency across a wide range of 3D tasks.

Multi-subject personalized image generation aims to synthesize customized images containing multiple specified subjects without requiring test-time optimization. However, achieving fine-grained independent control over multiple subjects remains challenging due to difficulties in preserving subject fidelity and preventing cross-subject attribute leakage. We present FocusDPO, a framework that adaptively identifies focus regions based on dynamic semantic correspondence and supervision image complexity. During training, our method progressively adjusts these focal areas across noise timesteps, implementing a weighted strategy that rewards information-rich patches while penalizing regions with low prediction confidence. The framework dynamically adjusts focus allocation during the DPO process according to the semantic complexity of reference images and establishes robust correspondence mappings between generated and reference subjects. Extensive experiments demonstrate that our method substantially enhances the performance of existing pre-trained personalized generation models, achieving state-of-the-art results on both single-subject and multi-subject personalized image synthesis benchmarks. Our method effectively mitigates attribute leakage while preserving superior subject fidelity across diverse generation scenarios, advancing the frontier of controllable multi-subject image synthesis.

Reconstructing realistic 3D human avatars from monocular videos is a challenging task due to the limited geometric information and complex non-rigid motion involved. We present MonoCloth, a new method for reconstructing and animating clothed human avatars from monocular videos. To overcome the limitations of monocular input, we introduce a part-based decomposition strategy that separates the avatar into body, face, hands, and clothing. This design reflects the varying levels of reconstruction difficulty and deformation complexity across these components. Specifically, we focus on detailed geometry recovery for the face and hands. For clothing, we propose a dedicated cloth simulation module that captures garment deformation using temporal motion cues and geometric constraints. Experimental results demonstrate that MonoCloth improves both visual reconstruction quality and animation realism compared to existing methods. Furthermore, thanks to its part-based design, MonoCloth also supports additional tasks such as clothing transfer, underscoring its versatility and practical utility.

Recent advancements in Audio-Video Large Language Models (AV-LLMs) have enhanced their capabilities in tasks like audio-visual question answering and multimodal dialog systems. Video and audio introduce an extended temporal dimension, resulting in a larger key-value (KV) cache compared to static image embedding. A naive optimization strategy is to selectively focus on and retain KV caches of audio or video based on task. However, in the experiment, we observed that the attention of AV-LLMs to various modalities in the high layers is not strictly dependent on the task. In higher layers, the attention of AV-LLMs shifts more towards the video modality. In addition, we also found that directly integrating temporal KV of audio and spatial-temporal KV of video may lead to information confusion and significant performance degradation of AV-LLMs. If audio and video are processed indiscriminately, it may also lead to excessive compression or reservation of a certain modality, thereby disrupting the alignment between modalities. To address these challenges, we propose AccKV, an Adaptive-Focusing and Cross-Calibration KV cache optimization framework designed specifically for efficient AV-LLMs inference. Our method is based on layer adaptive focusing technology, selectively focusing on key modalities according to the characteristics of different layers, and enhances the recognition of heavy hitter tokens through attention redistribution. In addition, we propose a Cross-Calibration technique that first integrates inefficient KV caches within the audio and video modalities, and then aligns low-priority modality with high-priority modality to selectively evict KV cache of low-priority modality. The experimental results show that AccKV can significantly improve the computational efficiency of AV-LLMs while maintaining accuracy.

We propose Anomagic, a zero-shot anomaly generation method that produces semantically coherent anomalies without requiring any exemplar anomalies. By unifying both visual and textual cues through a crossmodal prompt encoding scheme, Anomagic leverages rich contextual information to steer an inpainting‐based generation pipeline. A subsequent contrastive refinement strategy enforces precise alignment between synthesized anomalies and their masks, thereby bolstering downstream anomaly detection accuracy. To facilitate training, we introduce AnomVerse, a collection of 12,987 anomaly–mask–caption triplets assembled from 13 publicly available datasets, where captions are automatically generated by multimodal large language models using structured visual prompts and template‐based textual hints. Extensive experiments demonstrate that Anomagic trained on AnomVerse can synthesize more realistic and varied anomalies than prior methods, yielding superior improvements in downstream anomaly detection. Furthermore, Anomagic can generate anomalies for any normal‐category image using user‐defined prompts, establishing a versatile foundation model for anomaly generation.

Vision Language Models (VLMs) have shown strong performance in multimodal understanding, offering promise for the circuit-to-netlist translation task. However, the diverse component symbols and complex connections in circuit images challenge VLMs in understanding physical layouts and reasoning for electrical connection logic. To address these, we propose Circuit-Think, the first multimodal reasoning framework for the automated circuit-to-netlist translation task, which employs a Trajectory-Guided Reinforcement Learning (TGRL) paradigm for structured logical reasoning on circuit images. Circuit-Think initializes reasoning capabilities through supervised fine-tuning (SFT) on image-netlist pairs, then optimizes reasoning trajectories and netlist generation decisions using TGRL. Firstly, TGRL introduces a step-by-step reasoning paradigm, which guides the model with stepwise reward functions to simulate the human cognitive trajectory of ``identifying ports, recognizing devices, and inferring connections''. Secondly, we customize a multi-level reward that maps reasoning and answers into graph structures and node sets, jointly optimizing logical consistency and netlist accuracy via graph similarity and set matching. Thirdly, TGRL contains a reflective learning mechanism for low-scoring samples, which corrects the reasoning trajectory through reference answers as hints, avoiding local optima caused by sparse reward signals or erroneous reasoning paths. Moreover, we construct a circuit image-netlist reasoning dataset with 3,100 samples, offering step-by-step annotations for converting circuit images to netlists. Extensive experiments demonstrate that Circuit-Think achieves SOTA netlist accuracy and significantly improves the accuracy of downstream tasks.

Satire, a form of artistic expression combining humor with implicit critique, holds significant social value by illuminating societal issues. Despite its cultural and societal significance, satire comprehension, particularly in purely visual forms, remains a challenging task for current vision-language models. This task requires not only detecting satire but also deciphering its nuanced meaning and identifying the implicated entities. Existing models often fail to effectively integrate local entity relationships with global context, leading to misinterpretation, comprehension biases, and hallucinations. To address these limitations, we propose SatireDecoder, a training-free framework designed to enhance satirical image comprehension. Our approach proposes a multi-agent system performing visual cascaded decoupling to decompose images into fine-grained local and global semantic representations. In addition, we introduce a chain-of-thought reasoning strategy guided by uncertainty analysis, which breaks down the complex satire comprehension process into sequential subtasks with minimized uncertainty. Our method significantly improves interpretive accuracy while reducing hallucinations. Experimental results validate that SatireDecoder outperforms existing baselines in comprehending visual satire, offering a promising direction for vision-language reasoning in nuanced, high-level semantic tasks.

3D multi-human reconstruction from single images holds significant potential for advancing AR/VR applications. While remarkable progress has been made in single-human reconstruction, existing methods face challenges when reconstructing multiple humans. These challenges include: (1) severe inter-occlusion that disrupts individual body structures, and (2) the absence of physically plausible relative positioning among subjects. We present DECON, a novel DEcouple-and-reCONstruct framework that systematically addresses these limitations through two technical innovations: (1) a decouple-and-reconstruct framework with multi-view synthesis. It separates individuals and reconstructs detailed 3D bodies from a single image. (2) a Perspective-Aware Position Optimization (PAPO) approach. It ensures realistic positioning by fixing overlaps and gaps between subjects. Extensive experiments demonstrate our method's capability to reconstruct fully separated, anatomically complete 3D humans with clothed-geometric details and plausible interactions. Quantitative evaluations show a 54% reduction in Chamfer Distance and 35% in Point-to-Surface Distance compared to state-of-the-art methods.

Fine-Grained Image Retrieval~(FGIR) faces challenges in learning discriminative visual representations to retrieve images with similar fine-grained features. Current leading FGIR solutions typically follow two regimes: enforce pairwise similarity constraints in the semantic embedding space, or incorporate a localization sub-network to fine-tune the entire model. However, such two regimes tend to overfit the training data while forgetting the knowledge gained from large-scale pre-training, thus reducing their generalization ability. In this paper, we propose a Dual-Vision Adaptation (DVA) approach for FGIR, which guides the frozen pre-trained model to perform FGIR through collaborative sample and feature adaptation. Specifically, we design Object-Perceptual Adaptation, which modifies input samples to help the pre-trained model perceive critical objects and elements within objects that are helpful for category prediction. Meanwhile, we propose In-Context Adaptation, which introduces a small set of parameters for feature adaptation without modifying the pre-trained parameters. This makes the FGIR task using these adapted features closer to the task solved during the pre-training. Additionally, to balance retrieval efficiency and performance, we propose Discrimination Perception Transfer to transfer the discriminative knowledge in the object-perceptual adaptation to the image encoder using the knowledge distillation mechanism. Extensive experiments show that DVA performs well on three fine-grained datasets.

Reconstructing high dynamic range (HDR) images from low dynamic range (LDR) bursts plays an essential role in the computational photography. Impressive progress has been achieved by learning-based algorithms which require LDR-HDR image pairs. However, these pairs are hard to obtain, which motivates researchers to delve into the problem of annotation-efficient HDR image reconstructing: how to achieve comparable performance with limited HDR ground truths (GTs). This work attempts to address this problem from the view of semi-supervised learning where a teacher model generates pseudo HDR GTs for the LDR samples without GTs and a student model learns from pseudo GTs. Nevertheless, the confirmation bias, i.e., the student may learn from the artifacts in pseudo HDR GTs, presents an impediment. To remove this impediment, an uncertainty-based masking process is proposed to discard unreliable parts of pseudo GTs at both pixel and patch levels, then the student can learn from the trusted areas. With this novel masking process, our semi-supervised HDR reconstructing method not only outperforms previous annotation-efficient algorithms, but also achieves comparable performance with up-to-date fully-supervised methods by using only 6.7% HDR GTs.

Currently, pretrained models are rapidly scaling in size, which substantially increases the cost of fine-tuning them for downstream tasks. To address this challenge, parameter-efficient fine-tuning (PEFT) methods have been developed to optimize a minimal set of parameters for adaptation. While current PEFT approaches predominantly employ an "additive'' strategy, introducing learnable modules into inputs or architectures, neglect the inherent knowledge embedded within pretrained models, which may be redundant or even conflict with downstream tasks. This limitation leads to increased inference latency and suboptimal transfer performance, particularly in scenarios with significant domain gaps. In this paper, we propose a Subtractive Fine-tuning Paradigm(SFP), which converts multiple redundant operations within the original module into a linear transformation to enhance inference speed and model performance. Specifically, we introduce a compact filter block to replace specific module with interference and redundancy in the original structure to reduce model conflicts. By using a pseudo inverse matrix to construct filter block, ensuring that it can inherit the knowledge of the replacement module, and then freezing the rest of the model, only fine-tuning the filter block is performed to eliminate interference and redundant knowledge, thereby enhancing the model’s adaptability to downstream tasks. Experimental results demonstrate that our SFP outperforms existing PEFT methods in accuracy while decreasing the overall model parameters by 12%. Compared to full fine-tuning, the accuracy has increased by 8.47%(74.04% vs. 65.57%, VTAB).

Music-induced painting is a unique artistic practice, where visual artworks are created under the influence of music. Evaluating whether a painting faithfully reflects the music that inspired it poses a challenging perceptual assessment task. Existing methods primarily rely on emotion recognition models to assess the similarity between music and painting, but such models introduce considerable noise and overlook broader perceptual cues beyond emotion. To address these limitations, we propose a novel framework for music-induced painting assessment that directly models perceptual coherence between music and visual art. We introduce MPD, the first large-scale dataset of music–painting pairs annotated by domain experts based on perceptual coherence. To better handle ambiguous cases, we further collect pairwise preference annotations. Building on this dataset, we present MPJudge, a model that integrates music features into a visual encoder via a modulation-based fusion mechanism. To effectively learn from ambiguous cases, we adopt Direct Preference Optimization for training. Extensive experiments demonstrate that our method outperforms existing approaches. Qualitative results further show that our model more accurately identifies music-relevant regions in paintings.

Video transitions are critical for ensuring temporal coherence in edited media, yet existing methods often rely on handcrafted effects or relative-scale trajectories that fail to capture the physical structure of real-world scenes. In this work, we introduce a scale-aware video transition framework that explicitly incorporates depth-aware 3D reasoning into a diffusion-based generation pipeline. Built upon a powerful I2V foundation, our method leverages single-image depth prediction to align camera motion with metric-scale geometry, enabling physically consistent transitions. To reduce reliance on precise camera inputs, we propose a bidirectional conditional control module and a progressive training strategy with conditional dropout, enhancing generalization to loosely specified or missing camera trajectories. Extensive experiments demonstrate that our approach achieves state-of-the-art performance, delivering realistic, geometrically coherent transitions across diverse scenes and applications with minimal input guidance.

Remote sensing images are becoming increasingly widespread in military, earth resource exploration. Because of the limitation of a single sensor, we can obtain high spatial resolution grayscale panchromatic (PAN) images and low spatial resolution color multispectral (MS) images. Therefore, an important issue is to obtain a color image with high spatial resolution when there is only a PAN image at the input. The existing methods improve spatial resolution using super-resolution (SR) technology and spectral recovery using colorization technology. However, the SR technique cannot improve the spectral resolution, and the colorization technique cannot improve the spatial resolution. Moreover, the pansharpening method needs two registered inputs and can not achieve SR. As a result, an integrated approach is expected. We designed a novel multi-function model (MFmamba) to realize the tasks of SR, spectral recovery, joint SR and spectral recovery through three different inputs. Firstly, MFmamba utilizes UNet++ as the backbone, and a Mamba Upsample Block (MUB) is combined with UNet++. Secondly, a Dual Pool Attention (DPA) is designed to replace the skip connection in UNet++. Finally, a Multi-scale Hybrid Cross Block (MHCB) is proposed for initial feature extraction. Many experiments show that MFmamba is competitive in evaluation metrics and visual results and performs well in the three tasks when only the input PAN image is used.

Text-guided image inpainting aims to inpaint masked image regions based on a textual prompt while preserving the background. Although diffusion-based methods have become dominant, their property of modeling the entire image in latent space makes it challenging for the results to align well with prompt details and maintain a consistent background. To address these issues, we explore Mask AutoRegressive (MAR) models for this task. MAR naturally supports image inpainting by generating latent tokens corresponding to mask regions, enabling better local controllability without altering the background. However, directly applying MAR to this task makes the inpainting content either ignore the prompts or be disharmonious with the background context. Through analysis of the attention maps from the inpainting images, we identify the impact of background tokens on text tokens during the MAR generation, and leverage this to designToken Painter, a training-free text-guided image inpainting method based on MAR. Our approach introduces two key components: (1) Dual-Stream Encoder Information Fusion (DEIF), which fuses the semantic and context information from text and background in frequency domain to produce novel guidance tokens, allowing MAR to generate text-faithful inpainting content while keeping harmonious with background context. (2) Adaptive Decoder Attention Score Enhancing (ADAE), which adaptively enhances attention scores on guidance tokens and inpainting tokens to further enhance the alignment of prompt details and the content visual quality. Extensive experiments demonstrate that our training-free method outperforms prior state-of-the-art methods across almost all metrics.

Inspired by Segment Anything 2, which generalizes segmentation from images to videos, we propose SAM2MOT—a novel segmentation-driven paradigm for multi-object tracking that breaks away from the conventional detection-association framework. In contrast to previous approaches that treat segmentation as auxiliary information, SAM2MOT places it at the heart of the tracking process, systematically tackling challenges like false positives and occlusions. Its effectiveness has been thoroughly validated on major MOT benchmarks. Furthermore, SAM2MOT integrates pre-trained detector, pre-trained segmentor with tracking logic into a zero-shot MOT system that requires no fine-tuning. This significantly reduces dependence on labeled data and paves the way for transitioning MOT research from task-specific solutions to general-purpose systems. Experiments on DanceTrack, UAVDT, and BDD100K show state-of-the-art results. Notably, SAM2MOT outperforms existing methods on DanceTrack by +2.1 HOTA and +4.5 IDF1, highlighting its effectiveness in MOT.

Recent advancements in end-to-end autonomous driving systems (ADSs) underscore their potential for perception and planning capabilities. However, challenges remain. Complex driving scenarios contain rich semantic information, yet ambiguous or noisy semantics can compromise decision reliability, while interference between multiple driving tasks may hinder optimal planning. Furthermore, prolonged inference latency slows decision-making, increasing the risk of unsafe driving behaviors. To address these challenges, we propose ExpertAD, a novel framework that enhances the performance of ADS with Mixture of Experts (MoE) architecture. We introduce a Perception Adapter (PA) to amplify task-critical features, ensuring contextually relevant scene understanding, and a Mixture of Sparse Experts (MoSE) to minimize task interference during prediction, allowing for effective and efficient planning. Our experiments show that ExpertAD reduces average collision rates by up to 20% and inference latency by 25% compared to prior methods. We further evaluate its multi-skill planning capabilities in rare scenarios (e.g., accidents, yielding to emergency vehicles) and demonstrate strong generalization to unseen urban environments. Additionally, we present a case study that illustrates its decision-making process in complex driving scenarios.

Score Distillation Sampling (SDS) has achieved remarkable success in text-to-3D content generation. However, SDS-based methods struggle to maintain semantic fidelity for user prompts, particularly when involving multiple objects with intricate interactions. While existing approaches often address 3D consistency through multiview diffusion model fine-tuning on 3D datasets, this strategy inadvertently exacerbates text-3D alignment degradation. The limitation stems from SDS's inherent accumulation of view-independent biases during optimization, which progressively diverges from the ideal text alignment direction. To alleviate this limitation, we propose a novel SDS objective, dubbed as Textual Coherent Score Distillation (TCSD), which integrates alignment feedback from multimodal large language models (MLLMs). Our TCSD leverages cross-modal understanding capabilities of MLLMs to assess and guide the text-3D correspondence during the optimization. We further develop 3DLLaVA-CRITIC - a fine-tuned MLLM specialized for evaluating multiview text alignment in 3D generations. Additionally, we introduce an LLM-layout initialization that significantly accelerates optimization convergence through semantic-aware spatial configuration. Our framework, CoherenDream, achieves consistent improvement across multiple metrics on TIFA subset.As the first study to incorporate MLLMs into SDS optimization, we also conduct extensive ablation studies to explore optimal MLLM adaptations for 3D generation tasks.