While video-to-audio generation has achieved remarkable progress in semantic and temporal alignment, most existing studies focus solely on these aspects, paying limited attention to the spatial perception and immersive quality of the synthesized audio. This limitation stems largely from current models' reliance on mono audio datasets, which lack the binaural spatial information needed to learn visual-to-spatial audio mappings. To address this gap, we introduce two key contributions: we construct BinauralVGGSound, the first large-scale video-binaural audio dataset designed to support spatially aware video-to-audio generation; and we propose an end-to-end spatial audio generation framework guided by visual cues that explicitly models spatial features. Our framework incorporates a visual-guided audio spatialization module that ensures the generated audio exhibits realistic spatial attributes and layered spatial depth while maintaining semantic and temporal alignment. Experiments show that our approach substantially outperforms state-of-the-art models in spatial fidelity and delivers a more immersive auditory experience, without sacrificing temporal or semantic consistency. All datasets, code, and model checkpoints are available at: https://github.com/renlinjie868-web/SpatialV2A.
论文检索
输入标题、作者或关键词，从 2,006 篇学术成果中精准定位
While Vision-Language Models (VLMs) excel at semantic understanding, they struggle to comprehend 3D spatial relationships from limited views. Their reliance on implicit geometric encoding often leads to severe hallucinations and inconsistencies in spatial reasoning tasks. To address this, we introduce GeoMind, a model-then-reason framework that employs a single LLM to autoregressively generate an explicit Geometric Description Language (GDL) map, serving as a grounded context to derive the final answer. This intermediate GDL map provides an explicit and queryable world representation. Leveraging this explicit representation, we enforce a strict referential constraint, compelling the model to ground reasoning solely on the instantiated entities to ensure referential integrity and auditability. Specifically, we lift multi-view observations into object-centric tokens using frozen geometric priors and instance masks. The LLM is trained via a two-stage curriculum with programmatic supervision to generate the GDL map as a prerequisite for answering. On five spatial understanding benchmarks in both image and video settings, GeoMind delivers average accuracy gains of +6.9% (2B) and +9.8% (8B) over Qwen3-VL baselines. Our results suggest that explicit geometric grounding enables robust spatial reasoning without human annotation, providing a scalable and practical route to stronger spatial intelligence in large VLMs.
High dynamic range (HDR) imaging is crucial for capturing real-world lighting conditions. HDR imaging is traditionally achieved either by fusing multiple exposure frames or via inverse tone mapping from a single SDR image. However, the multi-exposure HDR method is prone to motion-induced artefacts and imposes demanding hardware requirements, limiting its practical applicability. Traditional inverse tone-mapping techniques primarily rely on pixel-wise regression methods, which ignore semantic scene contexts and thus frequently introduce halo artefacts and structural distortions. To address this limitation, this paper proposes MLLM-ITM, a novel inverse tone-mapping framework incorporating multimodal large language models (MLLMs). MLLM-ITM utilizes cross-modal features extracted from a frozen MLLM to simultaneously encode visual features and semantic understanding. These features are integrated into a downstream HDR reconstruction backbone through lightweight adapters, enabling content-aware dynamic-range expansion. Furthermore, the decoupled design between MLLM and HDR backbone avoids costly fine-tuning of MLLMs and remains model-agnostic, allowing effortless substitution with emerging multimodal architectures. Extensive experiments on public benchmarks demonstrate that the proposed MLLM-ITM achieves state-of-the-art performance compared with existing inverse tone-mapping methods, highlighting the effectiveness of cross-modal semantic priors in enhancing HDR imaging performance.
Accurate identification of non-rigid motion is crucial for geometric validity in self-supervised monocular depth estimation (MDE), yet it remains challenging for current methods. Inspired by Human Visual Perception (HVP), we present VFM-Dynamo, an efficient self-supervised MDE framework that disambiguates non-rigid objects by combining a vision foundation model (VFM) such as Grounding DINO and SAM with coarse to fine motion identification strategy: 1) Integrating visual-textual feature dependencies, we activate the capability of Grounding DINO global understanding with text prompts to detect potential object bounding boxes. 2) We then estimate coarse motion mask from reprojection error and photometric consistency to separate static from dynamic content. This mask provides coarse guidance for constraining rigid flow in static regions through two analogous estimates aligned by a consistency loss, and it drives a non-dynamical suppression module that removes boxes associated with static objects. 3) The remaining boxes are used to prompt SAM to produce refined motion identification, which in turn guide the joint optimization of depth and optical flow. Additionally, due to foreground–background mixing, standard interpolation-based upsampling often produces boundary artifacts. We introduce a learnable neighbor affinity interpolation (LNAI) module, which directly upsamples depth to full resolution and can be seamlessly integrated as a plug-and-play component. Experiments on a series of benchmarks demonstrate that proposed framework achieves state-of-the-art performance among unsupervised methods. Code is available at https://github.com/Qianqian3764/VFM-Dynamo.
Multi-object tracking (MOT) has long faced challenges with identity switches, especially for targets with low appearance discriminability and complex motion. Existing end-to-end trackers typically enhance robustness by modeling long-term temporal information across target-level representations, yet these representations remain insufficiently discriminative for spatially and visually similar targets. In this paper, we present MoTRa, a Motion-aware Target Representation learning framework for end-to-end MOT that enriches target-level representations with adaptive motion cues. As a result, each representation adaptively balances motion cues and appearance-related content cues under varying conditions, enhancing its discriminability for tracking. To achieve this, we propose a feature fusion and alignment module that extracts motion cues and adaptively fuses them with content features into target representations. To further regularize the dynamically fused representations, we propose a target-specific contrastive learning strategy that promotes intra-trajectory consistency and inter-target separability. Experimental results demonstrate the effectiveness of our method, achieving competitive performance on multiple benchmarks.
Unified remote sensing multimodal models exhibit a pronounced spatial reversal curse: although they can accurately recognize and describe object locations in images, they often fail to faithfully execute the same spatial relations during text-to-image generation, where such relations constitute core semantic information in remote sensing. Motivated by this observation, we propose Uni-RS, the first unified multimodal model tailored for remote sensing, to explicitly address the spatial asymmetry between understanding and generation. Specifically, we first introduce explicit Spatial-Layout Planning to transform textual instructions into spatial layout plans, decoupling geometric planning from visual synthesis. We then impose Spatial-Aware Query Supervision to bias learnable queries toward spatial relations explicitly specified in the instruction. Finally, we develop Image–Caption Spatial Layout Variation to expose the model to systematic geometry-consistent spatial transformations. Extensive experiments across multiple benchmarks show that our approach substantially improves spatial faithfulness in text-to-image generation, while maintaining strong performance on multimodal understanding tasks like image captioning, visual grounding, and VQA tasks.
The rapid advancement of AIGC video generation calls for evaluation frameworks that move beyond technical fidelity and incorporate human-centered aesthetic assessment. Existing benchmarks often overlook fine-grained perceptual qualities such as visual aesthetics, artistic style, and human preference. To address this limitation, we introduce VGA-BenchV2, an extended human-aligned benchmark and optimization framework for jointly evaluating and improving video generation quality and aesthetic value. Built upon VGA-Bench, VGA-BenchV2 preserves the original fine-grained taxonomy with two primary dimensions—Aesthetic and Generation—and 52 sub-dimensions. Guided by this taxonomy, we curate 1,016 diverse prompts and collect over 60,000 videos generated by 12 mainstream video generation models. More importantly, VGA-BenchV2 substantially expands human-labeled supervision by adding 36,000 task-level annotations, including 16,200 for aesthetic quality, 13,200 for aesthetic tagging, and 6,600 for generation quality, corresponding to 13.46×, 11.15×, and 1.55× scale-ups over VGA-Bench, respectively. Leveraging this enlarged annotation corpus, we develop a hybrid evaluator architecture consisting of VAQA-Net for continuous aesthetic scoring and two Qwen-based Large Vision-Language Model evaluators, VTag-Net and VGQA-Net, for aesthetic tagging and generation quality assessment. Extensive experiments demonstrate strong alignment with human judgments across diverse generation models. Beyond evaluation, VGA-BenchV2 further introduces an evaluation-to-optimization pipeline, where the learned aesthetic evaluator serves as a reward model for reinforcement learning-based generator fine-tuning. This closes the loop from benchmark construction and human supervision to automated evaluation and model optimization, enabling video generators to improve not only in realism but also in aesthetic quality and human preference alignment. Resources are available at https://huggingface.co/datasets/BestiVictoryLab/VGA-Bench.
Concept erasure aims to suppress sensitive content in diffusion models, but recent studies show that erased concepts can still be reawakened, revealing vulnerabilities in erasure methods. Existing reawakening methods mainly rely on prompt-level optimization to manipulate sampling trajectories, neglecting other generative factors, which limits a comprehensive understanding of the underlying dynamics. In this paper, we model the generation process as an implicit function to enable a comprehensive theoretical analysis of multiple factors, including textual conditions, model parameters, and latent states. We theoretically show that perturbing each factor can reawaken erased concepts. Building on this insight, we propose a novel concept reawakening method: Latent space Unblocking for concept REawakening (LURE), which reawakens erased concepts by reconstructing the latent space and guiding the sampling trajectory. Specifically, our semantic re-binding mechanism reconstructs the latent space by aligning denoising predictions with target distributions to reestablish severed text-visual associations. However, in multi-concept scenarios, naive reconstruction can cause gradient conflicts and feature entanglement. To address this, we introduce Gradient Field Orthogonalization, which enforces feature orthogonality to prevent mutual interference. Additionally, our Latent Semantic Identification-Guided Sampling (LSIS) ensures stability of the reawakening process via posterior density verification. Extensive experiments demonstrate that LURE enables simultaneous, high-fidelity reawakening of multiple erased concepts across diverse erasure tasks and methods.
Video captioning models convert frames into visual tokens and generate descriptions with large language models (LLMs). Since encoding all frames is prohibitively expensive, uniform sampling is the default choice, but it enforces equal temporal coverage while ignoring the uneven events distribution. This motivates a Learnable Frame Selector (LFS) that selects temporally diverse and event-relevant frames. LFS explicitly models temporal importance to balance temporal diversity and event relevance, and employs a stratified strategy to ensure temporal coverage while avoiding clustering. Crucially, LFS leverages caption feedback from frozen video-LLMs to learn frame selection that directly optimizes downstream caption quality. Additionally, we identify the gap between existing benchmark and human's cognition. Thus, we introduce ICH-CC built from carefully designed questions by annotators that reflect human-consistent understanding of video. Experiments indicate that LFS consistently improves detailed video captioning across two representative community benchmarks and ICH-CC, achieving up to 2.0% gains on VDC and over 4% gains on ICH-CC. Moreover, we observe that enhanced captions with LFS leads to improved performance on video question answering. Overall, LFS provides an effective and easy-to-integrate solution for detailed video captioning.
Driver cognitive distraction is a major cause of road collisions and remains difficult to detect. Unlike manual or visual distraction, cognitive distraction is diverted by thoughts unrelated to driving, even when the driver appears visually attentive and exhibits no explicit physical movements. In this work, we propose EyeCue, a gaze-empowered egocentric video understanding framework, to detect driver cognitive distraction. A key insight is that cognitive distraction manifests in the interaction between eye gaze and visual context. To capture this interaction, EyeCue integrates eye gaze with egocentric video to enable context-aware modeling of the driver's attention over time. Furthermore, to tackle the limited scale and diversity of existing datasets, we introduce CogDrive, a comprehensive multi-scenario dataset that augments four existing driving datasets with cognitive distraction annotations. Through extensive evaluations on CogDrive, we show that EyeCue achieves the highest accuracy of 74.38%, outperforming 11 baselines from 6 model families by over 7%. Notably, EyeCue can achieve an accuracy of over 70% across various driving scenarios (different road types, times of day, and weather conditions) with strong generalizability. These results highlight the importance of modeling gaze-context interactions and the effectiveness of cross-modal interaction modeling for multimodal cognitive distraction detection. Our codes and CogDrive dataset resources are available at https://github.com/langzhang2000/EyeCue.
Open‑vocabulary video instance segmentation (OV‑VIS) couples spatial‑temporal reasoning with language grounding, yet its adversarial robustness has remained unexplored. We present the Dual-Objective Triggers (DOT), the first transferable attack on OV-VIS that simultaneously exploits the vision–language coupling and temporal coherence. DOT deploys a Dual Semantic Perturbation Module that overlays two complementary triggers: a Semantic Suppression Trigger erases the alignment between the true object and the query, while a Plausible Replacement Trigger steers the tracker toward a phantom trajectory that is visually plausible and text‑consistent. To amplify cross‑model transferability without sacrificing perceptual fidelity, we introduce Phase‑Guided Adversarial Training, which injects perturbations primarily in the phase spectrum while blending amplitudes with clean references. Extensive experiments on four state‑of‑the‑art OV‑VIS implementations demonstrate that DOT reduces mAP by up to 69.3% and raises attack success rate by up to 98%, outperforming the strongest baselines by a factor of 1.6× on average, while maintaining a PSNR of 52.49 dB, thus exposing critical security vulnerabilities and laying a foundation for future research on robust and trustworthy vision–language systems.
While recent video generative models can synthesize high-fidelity videos, they struggle to portray plausible physical interactions and the resulting state transitions, a critical bottleneck for applications in robotics and VR/AR. To address this, we introduce a framework to generate a scalable synthetic dataset of controllable interactions. Our pipeline leverages a structured taxonomy and state-of-the-art image editing models to create explicit `start' and `end' state images, which serve as visual anchors for the interaction. To generate a seamless video utilizing these anchors, we propose State-Guided Sampling (SGS), a novel sampling technique that mitigates artifacts common in naive conditional generation. Furthermore, we develop and validate a new automated evaluation system that aligns with human judgments to ensure data quality. Experiments show that fine-tuning a base model on our dataset significantly enhances its ability to generate plausible interactions.
AI-driven education platforms have made some progress in personalisation, yet most remain constrained to static adaptation—predefined quizzes, uniform pacing, or generic feedback—limiting their ability to respond to learners’ evolving understanding. This shortfall highlights the need for systems that are both context-aware and adaptive in real time. We introduce PAL (Personal Adaptive Learner), an AI-powered platform that transforms lecture videos into interactive learning experiences. PAL continuously analyzes multimodal lecture content and dynamically engages learners through questions of varying difficulty, adjusting to their responses as the lesson unfolds. At the end of a session, PAL generates a personalized summary that reinforces key concepts while tailoring examples to the learner’s interests. By uniting multimodal content analysis with adaptive decision-making, PAL contributes a novel framework for responsive digital learning. Our work demonstrates how AI can move beyond static personalization toward real-time, individualized support, addressing a core challenge in AI-enabled education.
HARK: Hierarchical Agentic Retrieval with Keyframing for Video Understanding (Student Abstract)
PDF ↗Current video understanding models struggle with temporal reasoning and efficient processing while balancing detail preservation with computational efficiency. We propose a hierarchical memory system that segments videos into action and scene units, combined with question-aware agentic keyframe selection. Our method achieves 70.3% overall accuracy on VideoMME short video benchmarks.
Understanding human actions in videos requires more than raw pixel analysis; it relies on high-level semantic reasoning and effective integration of multimodal features. We propose a deep translational action recognition framework that enhances recognition accuracy by jointly predicting action concepts and auxiliary features from RGB video frames. At test time, hallucination streams infer missing cues, enriching feature representations without increasing computational overhead. To focus on action-relevant regions beyond raw pixels, we introduce two novel domain-specific descriptors. Object Detection Features (ODF) aggregate outputs from multiple object detectors to capture contextual cues, while Saliency Detection Features (SDF) highlight spatial and intensity patterns crucial for action recognition. Our framework seamlessly integrates these descriptors with auxiliary modalities such as optical flow, Improved Dense Trajectories, skeleton data, and audio cues. It remains compatible with state-of-the-art architectures, including I3D, AssembleNet, Video Transformer Network, FASTER, and recent models like VideoMAE V2 and InternVideo2. To handle uncertainty in auxiliary features, we incorporate aleatoric uncertainty modeling in the hallucination step and introduce a robust loss function to mitigate feature noise. Our multimodal self-supervised action recognition framework achieves state-of-the-art performance on multiple benchmarks, including Kinetics-400, Kinetics-600, and Something-Something V2, demonstrating its effectiveness in capturing fine-grained action dynamics.
Humans easily apply learned skills to different situations, a flexibility that AI systems still struggle to achieve. Current AI models are often confined to their training setup, leading to isolated developments and a narrow scope of application. This largely restricts the creation of flexible and general-purpose AI systems. Deep Model Reuse presents a novel solution. Imagine tapping into a vast library of pre-trained models, each a master in its specialized domain. Our approach re-purposes these existing models, extracting and transforming their knowledge for the development of novel AI systems. In this talk, we explore the essential techniques of this transformative process, highlighting the shift towards versatile and efficient AI that mirrors human cognition's adaptability. We introduce three foundational pillars of deep model reuse: understanding, composing, and refining. First, we investigate the internal behavior of neural networks—using language models as explainers and analyzing the representation space of diffusion models—to uncover how and what models have learned. Second, we develop methods to transform and compose models through weight mapping, knowledge distillation, and model dissection, enabling the creation of new capabilities by reassembling existing expertise. Third, we enhance reliability by editing model behaviors and mitigating biases, ensuring robustness in complex and dynamic environments. We demonstrate the power of this paradigm in generative AI, where model reuse leads to efficient diffusion models free from spectral bias, improved compositional understanding in video generation, and the repurposing of 2D/3D models for 3D/4D content creation. By shifting from training from scratch to intelligently reusing and recombining models, we move closer to adaptive, scalable, and human-like AI systems—ushering in a new era of sustainable and general intelligence.
Climate change poses a global threat to public health, food security, and economic stability. Addressing it requires evidence-based policies and a nuanced understanding of how the threat is perceived by the public, particularly within visual social media, where narratives quickly evolve through voices of individuals, politicians, NGOs, and institutions. This study investigates climate-related discourse on YouTube within the Brazilian context, a geopolitically significant nation in global environmental negotiations. Through three case studies, we examine (1) which psychological content traits most effectively drive audience engagement, (2) the extent to which these traits influence content popularity, and (3) whether such insights can inform the design of persuasive synthetic campaigns such as climate denialism using recent generative language models. Another contribution of this work is the release of a large publicly available dataset of 226K Brazilian YouTube videos and 2.7M user comments on climate change. The dataset includes fine-grained annotations of persuasive strategies, theory of mind categorizations in user responses, and typologies of content creators. This resource can help support future research on digital climate communication and the ethical risk of algorithmically amplified narratives and generative media.
Recent advancements in multimodal large language models for video understanding (videoLLMs) have enhanced their capacity to process complex spatiotemporal data. However, challenges such as factual inaccuracies, harmful content, biases, hallucinations, and privacy risks compromise their reliability. This study introduces Trust-videoLLMs, a first comprehensive benchmark evaluating 23 state-of-the-art videoLLMs (5 commercial, 18 open-source) across five critical dimensions: truthfulness, robustness, safety, fairness, and privacy. Comprising 30 tasks with adapted, synthetic, and annotated videos, the framework assesses spatiotemporal risks, temporal consistency and cross-modal impact. Results reveal significant limitations in dynamic scene comprehension, cross-modal perturbation resilience and real-world risk mitigation. While open-source models occasionally outperform, proprietary models generally exhibit superior credibility, though scaling does not consistently improve performance. These findings underscore the need for enhanced training datat diversity and robust multimodal alignment. Trust-videoLLMs provides a publicly available, extensible toolkit for standardized trustworthiness assessments, addressing the critical gap between accuracy-focused benchmarks and demands for robustness, safety, fairness, and privacy.
EchoBat: Echo-Vision Enhancement and Echo-Layered Sampling for Video LLMs Hallucination Mitigation
PDF ↗Recent advancements in multimodal large language models (MLLMs) have shown remarkable progress in video understanding. However, video MLLMs (VideoMLLMs) still suffer from hallucinations, generating nonsensical or irrelevant content. This issue partly stems from over-reliance on pre-trained knowledge, sometimes neglecting the rich visual information present in the video. Additionally, many existing methods rely on uniform frame sampling, which can overlook critical visual cues. To address these challenges, we present EchoBat, a novel approach that leverages audio information as well as video temporal and logical consistency to improve preference data construction and keyframe extraction. Our method integrates Direct Preference Optimization (DPO) to mitigate hallucinations by leveraging high-quality, contextually rich preference feedback. Specifically, we use GPT-4o to generate high-quality video descriptions and integrate visually relevant segments from Whisper-derived transcripts to construct preference responses. Correspondingly, we use the reference model itself to describe the reversed video, and use GPT-4o to flashback the text and fill in the hallucination to produce non-preferred responses. This strategy enhances the model’s ability to better understand visual content and temporal, logical relationships within videos. Furthermore, we propose an echo-layered sampling strategy for keyframe extraction from videos, which can provide more precise visual supervision compared to uniform sampling. Experimental results on the three latest video hallucination benchmarks demonstrate the effectiveness of our approach.
UniCUE: Unified Recognition and Generation Framework for Chinese Cued Speech Video-to-Speech Generation
PDF ↗Cued Speech (CS) enhances lipreading via hand coding, offering visual phonemic cues that support precise speech perception for the hearing-impaired. The task of CS Video-to-Speech generation (CSV2S) aims to convert CS videos into intelligible speech signals. Most existing research focuses on CS Recognition (CSR), which transcribes video content into text. Consequently, a common solution for CSV2S is to integrate CSR with a text-to-speech (TTS) system. However, this pipeline relies on text as an intermediate medium, which may lead to error propagation and temporal misalignment between speech and CS video dynamics. In contrast, directly generating audio speech from CS video (direct CSV2S) often suffer from the inherent multimodal complexity and the limited availability of CS data. To address these challenges, we propose UniCUE, the first unified framework for CSV2S that directly generates speech from CS videos without relying on intermediate text. The core innovation of UniCUE lies in integrating a understanding task (CSR) that provides fine-grained CS visual-semantic cues to to guide the speech generation. Specifically, UniCUE incorporates a pose-aware visual processor, a semantic alignment pool that enables precise visual–semantic mapping, and a VisioPhonetic adapter to bridge the understanding and generation tasks within a unified architecture. To support this framework, we construct UniCUE-HI, a large-scale Mandarin CS dataset containing 11,282 videos from 14 cuers, including both hearing-impaired and normal-hearing individuals. Extensive experiments conducted on this dataset demonstrate that UniCUE achieves state-of-the-art (SOTA) performance across multiple evaluation metrics.