Peer Revenue Publish or profit

Papers

Fresh from arXiv, typeset properly. Highlighted words trade on the market.

Computer Science Multimedia

cs.MM Watch Multimedia

Wednesday 19 August

Multimedia

GRACE: Grounded Reasoning via Adapter Composition and Evidence-Aware Calibration for Educational Visual Question Answering

Xinjin Li, Yudi Xia, Xi Zhao et al. (+8 more)

Educational visual question answering, or VQA, requires models to solve curriculum-oriented multiple-choice questions using both language and visual evidence. Compared with conventional open-ended VQA, educational examples often include structured assessment metadata, diagrams or image contexts, and semantically close answer options, creating strong opportunities for question-option shortcuts. We develop and evaluate a parameter-efficient adaptation framework for a frozen multimodal large language model in this setting. We introduce GRACE, Grounded Reasoning via Adapter Composition and Evidence-Aware Calibration, a framework that uses the pedagogical state of each question to specialize lightweight language and vision adaptation. The state combines inference-visible subject, grouped skill, grade, visual-context, question-intent, and option-structure cues. GRACE uses factor-specific prompts and lightweight visual adapters, then applies evidence-aware option calibration to score all candidates under a shared multimodal context. On ScienceQA, GRACE improves a shared-adapter baseline from 90.5 percent to 93.1 percent overall accuracy and from 88.7 percent to 91.2 percent on image-context questions. Removing pedagogical composition, option calibration, or the visual adapter reduces overall accuracy by 1.4, 1.0, and 1.5 points, respectively. These controlled results show that structured educational state is an effective routing signal for parameter-efficient multimodal adaptation.
Read the full abstract

All authors Xinjin Li, Yudi Xia, Xi Zhao, Yiliu Xu, Yining Liu, Cheng Lu, Yujian Long, Yu Ma, Jinghan Cao, Liang Fan, Yeyun Xu

Multimedia · Cross-listed from cs.LG

Pretraining Reusable Inference Across Views with Synthetic Task Priors

Jielong Lu, Zhihao Wu, Jiajun Yu et al. (+2 more)

Modern pretrained encoders make representations from heterogeneous views increasingly reusable, but the procedure that determines view utility and combines evidence is still relearned for each downstream task. Consequently, knowledge about view relevance, complementarity, reliability, and missingness is repeatedly discarded rather than transferred across tasks. We therefore reformulate multi-view learning as learning a reusable, task-conditioned inference procedure rather than a fixed fusion function. Based on this perspective, we propose SIMPLE, a prior-fitted multi-view in-context learner that predicts query labels by conditioning on a small labeled support set. Since existing real-world datasets cover only a limited range of view configurations and task structures, we construct a controllable synthetic task prior in embedding space. It generates diverse support-query episodes with varying class structures, shared and view-specific factors, representation geometries, cross-view dependencies, reliability levels, missingness patterns, and distribution shifts. A hierarchical inference architecture then performs reasoning within views, across views, and across support and query samples. Experiments on multi-view and multi-omics benchmarks demonstrate that the frozen variant of SIMPLE achieves competitive performance without updating the inference backbone, while lightweight adapter calibration attains leading performance on most evaluated datasets. Together, the results under frozen, one-shot, and missing-view settings support the central hypothesis that multi-view reasoning itself can be pretrained and reused, while lightweight adapter calibration provides task-specific alignment when needed.
Read the full abstract

All authors Jielong Lu, Zhihao Wu, Jiajun Yu, Zhaoliang Chen, Haishuai Wang

Multimedia · Cross-listed from cs.DL

MediaGraph: A Content-Aware Data Model and Query Framework for Multimodal Knowledge Graphs

Florian Ruosch and Luca Rossetto

Multimodal knowledge graphs typically treat multimedia documents as opaque, external entities. This content-agnostic approach constrains retrieval and analysis by isolating media from the graph's core structure, hindering the ability to capture and query complex relationships across media types. To address this, we introduce the MediaGraph Data Model and its prototypical implementation MeGraS, the MediaGraph Store, a novel approach that integrates multimedia content as graph nodes. This paradigm shift enables the query engine to directly access and process a document's intrinsic content, allowing for native operations such as feature-based similarity search, dynamic segmentation, and the inference of non-materialized relations. By extending the SPARQL query language, MeGraS provides a cohesive platform for the storage, management, and expressive querying of multimodal data. MeGraS is open-source software that establishes a new framework, moving the field toward content-aware multimodal knowledge graphs.
Read the full abstract

Tuesday 18 August

Multimedia · Cross-listed from cs.CL

SpeechSense: A Paralinguistic-Focused Dataset for Fine-Grained Speech Sentiment Analysis

Shicheng Ma, Wenqian Cui and Irwin King

Recent advances in AI have revolutionized speech processing, yet effective speech understanding requires discerning not just what is said, but how it is said. Speech Sentiment Analysis plays a critical role in decoding these paralinguistic cues for diverse real-world applications such as recruitment and customer service. However, existing Speech Sentiment Analysis research faces two primary limitations. First, dominant approaches rely on text-centric pipelines that cascade Automatic Speech Recognition with text analysis. This process inevitably discards essential acoustic features like prosody and tone, failing to capture attitudinal meanings in acoustically ambiguous utterances. Second, current benchmarks suffer from a mismatch in label granularity, prioritizing basic emotions (e.g., happy, sad) over the nuanced interpersonal stances (e.g., confident, impatient) necessary for social sensitivity. To address these limitations, we propose a novel dataset, SpeechSense, for fine-grained speech sentiment analysis. Specifically, we define a specialized 8-class taxonomy of interpersonal stances detectable primarily through prosodic cues beyond lexical content alone. We then construct a curated dataset based on this taxonomy, built from high-fidelity speech synthesis and rigorous human validation. Comprehensive experiments across multi-modal LLMs, text-only LLMs, and speech encoders demonstrate that models with acoustic access consistently outperform text-only baselines. These results empirically validate the primacy of acoustic cues in detecting subtle speaker attitudes, highlighting the necessity of SpeechSense. Dataset and supplementary materials are available at https://github.com/Sher13cked/SpeechSense.
Read the full abstract

Multimedia · Cross-listed from cs.SD

UniVerse: Benchmarking and Enhancing LALMs on Culturally Inclusive Low-Resource Music Understanding

Ziya Zhou, Shangda Wu, Shenyang Xu et al. (+16 more)

Recent advances in large audio-language models (LALMs) have significantly improved performance in tasks such as music captioning, genre classification, and sound event detection. However, limited attention has been paid to improving their adaptability across diverse musical traditions, particularly folk music rooted in distinct cultural contexts. Folk-music traditions are typically resource-scarce, unevenly represented across regions, and poorly documented. Even when such samples appear in large-scale pre-training, LALMs often fail to capture their structural and stylistic characteristics, partly due to the absence of dedicated evaluation protocols and training solutions. To address these limitations, we introduce UniVerse, a reproducible solution for low-resource music understanding. Specifically, we propose UniVerseBench, a benchmark of 5,042 Q&A pairs across more than 38 cultural and linguistic entities, constructed via an expert-guided yet highly automated pipeline. In parallel, we construct a fully automated, model-generated multi-turn dialogue training dataset UniVerseSet. By training LALMs on UniVerseSet, we systematically adapt and investigate representative multimodal imbalance learning strategies across both dense and Mixture-of-Experts (MoE) architectures. Experimental results indicate that fully automated data curation combined with imbalance-aware training yields non-trivial improvements, but models still struggle to capture fine-grained acoustic features, indicating a gap between surface-level alignment and deep musical comprehension.
Read the full abstract

All authors Ziya Zhou, Shangda Wu, Shenyang Xu, Yutong Zheng, Dafang Liang, Suin Chung, Danbinaerin Han, Junyan Jiang, Yongyi Zang, Ruibin Yuan, Rongxiu Zhong, Shilei Zhang, Junlan Feng, Jinglei Liu, Haotian Zhou, Zijin Li, Dasaem Jeong, Wei Xue, Yike Guo

Multimedia

On computational approaches to Pop music culture

Arthur Flexer

This overview article presents arguments why the computational study of Pop music culture needs to be conducted in a multi-modal way beyond mere audio analysis, gives a survey of already published quantitative work on analyzing Pop music at scale, and discusses challenges and promising research avenues for future work. We argue that Pop music culture is a rich tapestry of audio, visual, textual and cultural connotations and relations which needs to be studied in an integrative way as a multi-modal socio-cultural phenomenon. What is needed is an approach which is reminiscent of "distant reading", i.e. algorithmic analysis of thousands of books as a research tool in digital humanities. In addition to listening to audio, algorithms need to view album artwork and music videos, to read meta-information, lyrics, music magazines and books. Our review of already available work on distant reading/listening/viewing and multi-modal combinations thereof reveals two major open issues: a scarcity of truly multi-modal approaches and questionable external validity rooted in sampling practices when building music corpora. In trying to overcome these shortcomings we sketch three exemplary avenues for future research on Pop music culture: charting the topic universe of music lyrics, providing an iconography of album cover art, tracking retro cycles in music's timeline.
Read the full abstract

Multimedia · Cross-listed from cs.CV

DynaForcing: Overcoming Dynamic Collapse in Self-Forcing Distillation for Streaming Avatar Generation

Yubo Huang, Sirui Zhao, Xinchen Yao et al. (+5 more)

Audio-driven avatar generation requires realistic lip-sync, expressive motion, and real-time streaming. Recent work achieves the latter via self-forcing with Distribution Matching Distillation (DMD), but this paradigm suffers from a critical failure that has not been systematically characterized: dynamic collapse, where the student model converges to a near-static optimum with high perceptual quality but severely suppressed temporal dynamics. We trace this to two causes: the reverse KL objective in DMD, which biases toward low-motion modes, and unanchored self-conditioning, which creates a feedback loop that amplifies collapse. This is especially harmful for avatars, where even subtle motion loss breaks lip-sync and expression. To address this, we propose DynaForcing, a training framework with three complementary strategies applied at different levels. Specifically, Hybrid Forcing anchors rollouts to ground-truth dynamics at the data level to break the feedback loop. Dynamics-Aware Reward Regularization introduces explicit motion rewards via the RL interpretation of DMD to counteract the reverse KL bias at the loss level. Reference Perturbation perturbs reference images to decouple identity from static details, forcing the model to rely on audio for motion at the conditioning level. We further introduce computation graph pruning and gradient replay, reducing the GPU footprint of self-forcing by over an order of magnitude. Experiments show that DynaForcing recovers dynamics to teacher-comparable levels (Dyn-Deg: 0.31 -> 0.73, Sync-C: 7.03 -> 7.68) while improving visual quality, resolving the quality-dynamics trade-off throughout training without early stopping.
Read the full abstract

All authors Yubo Huang, Sirui Zhao, Xinchen Yao, Zhengye Zhang, Jinyang Huang, Fengqi Cui, Shiwei Wu, Enhong Chen

Multimedia · Cross-listed from cs.CV

SE-MoLoRA: Shared-Expert LoRA Adapters for Domain-Specific Photographic Assessment

Bishwash Khanal, Anlan Zhang, Sasu Tarkoma et al. (+2 more)

Vision-language models can describe images fluently, but they often fail to provide actionable photographic critique because semantic content and aesthetic judgment remain entangled. We propose SE-MoLoRA, a modular parameter-efficient adaptation framework for domain-specific photographic assessment. The method separates general photographic knowledge from specialist residual judgments using an always-active shared LoRA expert and routed adapters for composition, lighting, and technical quality. A lightweight query router selects the relevant specialist, enabling targeted critique without training separate full models. A rank-64 shared adapter captures broad photographic vocabulary, while rank-32 specialists learn domain-specific residuals with an orthogonal regularization penalty that encourages disentangled representations. Training data is obtained by distilling the Reddit Photo Critique Dataset into domain-labeled critique samples. On held-out critique generation, SE-MoLoRA improves BERTScore-F1 from 0.2317 to 0.4215 over monolithic LoRA and is preferred in 84.6% of pairwise comparisons, while using fewer active parameters than separate specialist models. SVD-based ablation study shows that shared-specialist decomposition and orthogonal regularization reduce expert overlap. These results demonstrate that modular adaptation improves controllability and specificity in multimodal photographic critique.
Read the full abstract

All authors Bishwash Khanal, Anlan Zhang, Sasu Tarkoma, Tommi Mikkonen, Abhishek Kumar

Monday 17 August

Multimedia · Cross-listed from cs.CV

Boosting Knowledge-based Visual Question Answering with Structured Context Reasoning

Qiyou Liu, Yong Zhang, Jianjie Luo et al. (+2 more)

Knowledge-based Visual Question Answering aims to answer questions about an image by integrating external knowledge with visual and textual information. Recent approaches often rely on in-context learning to prompt Large Language Models (LLMs) with multimodal context in a zero-shot or few-shot manner. However, we observe that directly concatenating heterogeneous visual descriptions and retrieved knowledge into long, unstructured prompts often degrades reasoning performance, due to both excessive irrelevant context and the lack of explicit relational structure. In this paper, we propose an LLM-based Structured Context Reasoning (SCoRe) framework that infers both explicit and implicit relationships for prediction. SCoRe consists of three stages: Context Acquisition, which generates diverse visual notes and retrieves explicit knowledge via an efficient two-stage multimodal retrieval strategy; Context Selection, which filters relevant visual, explicit, and implicit knowledge using LLM-guided selection; and Context Compression, which performs Relational Logic Distillation (RLD) to transform raw text into explicit entity-relation triplets. These relational triplets serve as a concise and structured prompt for final answer prediction. Extensive experiments on the OK-VQA and A-OKVQA benchmarks demonstrate that SCoRe consistently outperforms state-of-the-art methods.
Read the full abstract

All authors Qiyou Liu, Yong Zhang, Jianjie Luo, Zhenguo Yang, Yi Yu

Multimedia · Cross-listed from cs.CV

Steering the Flow: Inverting Face Recognition Models via Gradient-Guided Flow Matching

Ye Lu, Shen Wang, Zhaoyang Zhang et al. (+4 more)

Model Inversion Attacks (MIAs) aim to reconstruct representative training samples of target identities from face recognition models, exposing critical security vulnerabilities. Existing methods typically rely on indirect guidance or highly stochastic guidance, making it difficult to stably optimize generation trajectories toward target facial images. In this paper, we propose Steering Flow Model Inversion (SFMI), a novel two-stage white-box model inversion method that reformulates inversion as a trajectory-steering task. Specifically, Step I, Learning a Generic Flow Matching Prior, pre-trains a generic unconditional Flow Matching model to encode the manifold of human faces as a robust prior. Step II, Attacking with Progressive Guidance Scheduler (PGS), injects time-dependent target-specific gradients during sampling. By backpropagating through the target model to obtain gradients from intermediate generated states, PGS progressively injects adaptive guidance signals into the vector field. This process effectively steers the current generative flow from random noise toward the high-density regions of the target class. Under an identity-disjoint cross-evaluation setting using the CelebA dataset, SFMI achieves an ACC of 0.9248, an FID of 22.61, and an LPIPS of 0.3874 on the ArcFace target. Extensive experiments on multiple target models demonstrate that SFMI achieves competitive state-of-the-art performance in attack success and visual fidelity under the evaluated white-box protocol.
Read the full abstract

All authors Ye Lu, Shen Wang, Zhaoyang Zhang, Yihan Yan, Li Liu, Runze Liu, Fanghui Sun

Multimedia · Cross-listed from cs.LG

UniTAC: Universal Task-Aware Compression via Weighted Distortion Measures

Homa Esfahanizadeh, Matin Mortaheb, Adeel Mahmood et al. (+2 more)

Lossy compression is conventionally driven by a task-agnostic distortion (e.g., MSE or MS-SSIM), yet in many emerging applications the receiver cares not about uniform fidelity but about a downstream task whose relevant content varies across the signal and evolves over time. We formulate task-aware compression as a weighted rate-distortion problem, in which a single codec is driven by a separable, per-component weighted distortion whose weights encode task importance and may depend on the source. We introduce task consistency, i.e., that minimizing the weighted distortion also minimizes the true task loss, and characterize when it holds: for linear tasks, the task loss admits a weighted-MSE form with signal-independent weights under suitable cross-term conditions, while for nonlinear tasks, an integrated-gradients analysis motivates separable task-aware weights. We show how task symmetry and irrelevance further constrain the admissible weights. Guided by this theory, we realize the weight-conditioned code in a single learned Vision Transformer (ViT) codec whose token-level attention natively consumes a per-component importance vector, so one fixed backbone is re-targeted at runtime, from universal (task-agnostic) to task-specialized operation, purely by swapping the injected weights, without retraining, while producing a single human-viewable reconstruction steered to the active task. On downstream face-analysis tasks, a single model reaches 91.4% accuracy at 0.034 bpp on a localized task, within 1.9% of a task-specific codec (93.3%) and well above a universal codec (76.9%). Such task-adaptive compression suits bandwidth-constrained perception systems, e.g., in Physical AI, where the active task drifts and per-task retraining is infeasible.
Read the full abstract

All authors Homa Esfahanizadeh, Matin Mortaheb, Adeel Mahmood, Jinfeng Du, Harish Viswanathan

Multimedia · Cross-listed from cs.CV

Matched Outcomes, Divergent Gaze: How Foveated MLLMs Search Compared to Humans

Mohamed Amine Kerkouri, Marouane Tliba, Aladine Chetouani et al. (+2 more)

Human visual search is serial: the fovea must land on a candidate to confirm it, and those landings form a scanpath. Whether multimodal large language models (MLLMs), given the same foveated input, search as humans do bears on their use as models of human vision and on attention-alignment scores. We compare three general-purpose MLLMs with human eye-movement scanpaths on goal-directed search (COCO-Search18), driving each model fixation by fixation through an identical, human-matched foveated view and assessing it along three axes: the decision of target presence, the efficiency of reaching the target, and the gaze process itself. The axes dissociate. On the decision and on target acquisition the models match or exceed humans, detecting present targets near ceiling and reaching them on the first saccade more often than people do. The gaze process is not human. Under the human-matched condition, all three share one signature: low-entropy, large-amplitude, self-consistent scanpaths that agree with themselves far more closely than two humans agree with each other. That is consistent with a single-pass, non-serial architecture rather than a limit of acuity. Matched retinal input reproduces where humans look but not how the looking unfolds in time, and no degradation regime recovers human-like search at human-like success. The gap sits on a process axis that answer-alignment and saliency metrics do not measure. Because they miss it, such metrics cannot certify human-like vision, and zero-shot models suit outcome and spatial questions but not temporal, process-level ones.
Read the full abstract

All authors Mohamed Amine Kerkouri, Marouane Tliba, Aladine Chetouani, Ulas Bagci, Alessandro Bruno

Multimedia · Cross-listed from cs.CV

KeyID: Decoupled Drafting and Keyframe Editing for Identity-Preserving Video Generation

Jianjie Luo, Yiming Zhong, Haoming Shen et al. (+2 more)

Identity-preserving video generation (IPVG) requires synthesizing videos that are faithful to both reference subjects and text prompts. Existing methods are often hindered by high tuning costs or limited input-level enhancements, struggling to maintain rigid identity consistency during complex, long-sequence actions. To address these limitations, we propose KeyID, a training-free IPVG framework that decouples the synthesis of video dynamics from the injection of identity. Specifically, KeyID comprises two components: (1) Reference-Aware Video Generation, which produces an identity-agnostic video draft aligned with multiple references, and (2) Identity-Preserved Keyframe Editing, which integrates the target identity via sparse keyframe correction and subsequent motion interpolation. By shifting from dense frame-level supervision to sparse keyframe-level refinement, KeyID effectively resolves the capacity conflict between prompt adherence and identity fidelity. Crucially, our modular design allows seamless extension to multi-subject references and complex sequential action generation without additional training. KeyID outperforms prior works and is validated by automatic and human evaluations on the official challenge benchmark, ultimately securing the runner-up position in the Track 2 (Sequential Action) of the ACM Multimedia 2026 IPVG Grand Challenge. Source code is available at https://github.com/WISLab-GDUT/KeyID.
Read the full abstract

All authors Jianjie Luo, Yiming Zhong, Haoming Shen, Yupeng Xiao, Zhenguo Yang

Multimedia · Cross-listed from cs.GR

AnyTalk: Speech Animation for Arbitrary Characters Leveraging a Video Generation Model

Kwan Yun, Serin Yoon, Sunjin Jung et al. (+3 more)

We present AnyTalk, a novel method for generating 3D speech animations for arbitrary characters without requiring any animation data. While existing audio-driven 3D speech animation methods rely on character-specific training data or laborious rigging/re-meshing, AnyTalk circumvents these limitations by leveraging recent video diffusion models trained on extensive video datasets. We first adapt a pre-trained video diffusion model to a target character through our Character-specific Fine-tuning (CsF) technique. By fine-tuning on rendered images of the 3D character paired with zeroed-out audio embeddings (representing "no motion"), we eliminate the need for animation data while preserving the motion prior of large-scale video diffusion model. We then uplift the resulting talking-head video into a 3D speech animation by estimating blendshape parameters through a proposed optimization process. AnyTalk enables lip-synced animations across diverse face meshes and blendshape configurations, significantly reducing manual effort and data requirements. We further enhance usability by distilling AnyTalk into a streamlined network, $\text{AnyTalk}_{RT}$, thereby enabling real-time performance. By leveraging talking-head video generation, our method broadens access to audio-driven speech animation technology for arbitrary characters. The code is publicly available at https://serin-yoon.github.io/projects/anytalk/.
Read the full abstract

All authors Kwan Yun, Serin Yoon, Sunjin Jung, Jung Eun Yoo, Inyup Lee, Junyong Noh

Sunday 16 August

Multimedia · Cross-listed from cs.CV

EditStream: A Unified Autoregressive Framework for Interactive Video Generation and Editing

Yuqian Zhou, Zhenghong Zhou, Zongze Wu et al. (+5 more)

Interactive video generation and editing are becoming increasingly important for creative design. In this report, we introduce EditStream: a unified framework for interactive video generation and editing. EditStream unifies multiple video creation and manipulation tasks within a single DiT-based model through flexible task-specific conditioning, and further transforms it into a fast, few-step autoregressive model for efficient streaming. It supports Text-to-Video, Image-to-Video, Video-to-Video, Editing Propagation, Reference-guided Video Editing, and Camera Pose Change, enabling flexible control over video generation, transformation, and editing within one system. To make the unified model practical for interactive use, we develop a two-stage distillation approach that combines Velocity Moment Matching (VMM) with autoregressive unrolling. VMM matches conditional velocity moments at student-reached intermediate states to preserve generation quality and motion, while unrolling exposes the student to its own autoregressive predictions to improve temporal stability. Together, they alleviate common challenges in few-step autoregressive video generation, including over-saturation, degraded motion, temporal instability, and complex training. EditStream provides a practical and scalable solution that bridges high-quality diffusion-based video models with interactive creative workflows.
Read the full abstract

All authors Yuqian Zhou, Zhenghong Zhou, Zongze Wu, Cameron Smith, Richard Zhang, Jiebo Luo, Eli Shechtman, Zhe Lin

Multimedia · Cross-listed from cs.CV

CLARA: Clip-Level Multimodal Alignment with VLM-Derived Rationales for Hateful Video Detection

Yuchen Zhang, Shuang Dai, Zeyu Fu et al. (+3 more)

Hateful video detection has become increasingly important with the rapid growth of video-centric social media platforms, given the serious risks that hate speech poses to both individual well-being and social cohesion. Compared with text or static multimodal content, hateful video detection remains underexplored and significantly more challenging, as hateful meaning often arises from complex interactions among multimodal cues, including speech, audio, and visual content. Moreover, such signals are often brief, implicit, and temporally dependent, making them difficult to capture using conventional video-level representations. In this work, we propose CLARA, a clip-level multimodal framework for hateful video detection. Instead of treating a video as a single instance, CLARA models it as a sequence of fine-grained clips, enabling more precise capture of temporally localized hateful signals. We introduce a Mixture-of-Experts clip encoder for adaptive multimodal alignment, a local-global segment contrastive objective to jointly model short-term cues and long-range temporal dependencies, and VLM-derived rationales integrated via a gated Transformer to provide high-level semantic guidance. Extensive experiments on three hateful video datasets demonstrate that CLARA consistently outperforms state-of-the-art methods. Further ablation studies and parameter analyses validate the effectiveness of each component.
Read the full abstract

All authors Yuchen Zhang, Shuang Dai, Zeyu Fu, Yunfei Long, Ravi Shekhar, Haralambos Mouratidis

Multimedia · Cross-listed from cs.CV

Beyond Visual CoT: Internalized Visual Thinking for Proactive Video Reasoning

Xiaoyu Zhu, Xinke Deng, Suresh Taddewadikar et al. (+4 more)

Multimodal large language models increasingly use visual chain-of-thought (Visual CoT) to reason about spatial, temporal, and embodied environments. By generating intermediate reasoning images, Visual CoT provides an intuitive mechanism for visual foresight but introduces substantial inference overhead, which is particularly problematic for proactive video reasoning. We ask whether models can learn to think visually during training while reasoning directly at inference. We introduce Internalized Visual Thinking (IVT), a post-training framework that jointly optimizes textual prediction and next-embedding prediction over unlabeled videos. Given a partially observed video, IVT predicts latent representations of future frames together with the target textual answer, encouraging the model to capture motion, object transitions, interactions, and latent intent. At inference, IVT generates the answer directly without synthesizing or re-encoding future frames. We conduct controlled studies across target representations, decoder designs, prediction horizons, data mixtures, training curricula, and predictive objectives. IVT improves over direct-answer fine-tuning on all six evaluation settings while retaining the same inference pathway. Compared with explicit Visual CoT, IVT achieves comparable or better performance and reduces average end-to-end latency by more than 5x. Together, our findings suggest that explicit pixel-space generation at inference time, as used in visual chain-of-thought, may not be necessary for effective proactive video reasoning. Predictive world modeling can be internalized during training to produce multimodal reasoners that are both more accurate and substantially more efficient.
Read the full abstract

All authors Xiaoyu Zhu, Xinke Deng, Suresh Taddewadikar, Arnab Kumar Mondal, Zhongyu Jiang, Ian Fasel, Joerg Liebelt

Multimedia · Cross-listed from cs.RO

Scaling Manual-Grounded Appliance Manipulation with Data Synthesis and Unified Planning

Yuxing Long, Lei Kang, Ziyan Yu et al. (+8 more)

Operating household appliances requires long-horizon planning that is state-dependent and robust to disturbances, yet existing large models fall short, as no sufficiently diverse, task-oriented dataset exists to support such planning. To bridge this gap, we propose MAGE, a scalable data synthesis pipeline that introduces a novel Hierarchical Appliance Graph (HAG) to automatically generate part grounding, long-horizon planning, and closed-loop recovery data from appliance manuals. With MAGE, we build UseAppliance, the first large-scale dataset for manual-grounded appliance manipulation planning, spanning 22 appliance categories with 89K+ part annotations, 53K+ manipulation tasks, and 33K+ closed-loop adjustment steps. Built on UseAppliance, we develop AppliancePlan, an end-to-end model for manual-grounded appliance manipulation planning. On RealAppliance-Bench, AppliancePlan with only 7B parameters achieves over 10x the best baseline on open-loop planning and consistently outperforms state-of-the-art models across all tasks. Real-robot experiments on six household appliances further confirm effective sim-to-real transfer, marking an important step toward general-purpose household robotics.
Read the full abstract

All authors Yuxing Long, Lei Kang, Ziyan Yu, Yuzheng Gao, Bin Cheng, Jiyao Zhang, Xiaoqi Li, Haolin Yang, Dongjiang Li, Hui Shen, Hao Dong

Multimedia · Cross-listed from eess.AS

CineDub: Scaling End-to-End Video Dubbing to Multi-Speaker Dialogues with Coherent Sound Effects

Yusheng Dai, Kangdi Wang, Baolong Gao et al. (+4 more)

Automatic video dubbing in the wild remains fundamentally limited by two competing constraints: hierarchical methods depend on brittle, multi-stage preprocessing pipelines that severely restrict data scalability and practical deployment, while holistic approaches operating on uncropped video suffer from weak temporal alignment and speaker-utterance ambiguity in multi-speaker settings. To overcome these limitations, we propose CineDub, a unified diffusion-based model that achieves precise multi-speaker dialogue dubbing directly from uncropped videos, without face cropping or speaker diarization. Central to our approach is the Implicitly-Coupled Holistic Conditioning (ICHC) paradigm, where holistic visual representations and a semantic-bundled transcription format are encoded independently, yet implicitly coupled through cross-modal training to resolve speaker ambiguity and enable precise multi-speaker multi-turn dialogue dubbing. Building on the unified temporal cues captured by holistic visual features, we further extend CineDub to joint speech and audio generation. We introduce an Ambient-to-Linguistic Curriculum Learning (ALC) to mitigate sub-task degradation, and a decoupled textual branch control mechanism to resolve cross-prompt interference during simultaneous generation. We also release two in-the-wild benchmarks, CineDub-Multi for multi-speaker dialogue dubbing and CineDub-SA for video-to-speech-and-audio (V2SA) generation, to enable evaluation under realistic conditions. Experiments show that CineDub achieves state-of-the-art results on established single-speaker dubbing and video-to-audio benchmarks while excelling in multi-speaker dialogue dubbing and acoustically coherent joint generation.
Read the full abstract

All authors Yusheng Dai, Kangdi Wang, Baolong Gao, Yuxuan Jiang, Weiqiang Wang, Qiuhong Ke, Jianfei Cai

Multimedia · Cross-listed from cs.SD

Adding Voice Cloning to Text-to-Audio-Video Models with a Single Zero-Initialised Layer

Ivan Mikheev, Viacheslav Vasilev, Anna Dmitrienko et al. (+4 more)

Text-to-audio-video (T2AV) generation models produce a video and its soundtrack from a textual description, but offer no control over whose voice speaks in the output. We show that a base T2AV model can be turned into a voice-cloning model by adding a single zero-initialized linear layer on top of its audio backbone, fine-tuning for a comparatively short training schedule, and conditioning on a short reference recording at inference time. The reference is injected through two complementary signals: its diffusion latents are prepended to the audio stream, and a global speaker embedding modulates token of the target audio. On a benchmark of 674 speaker-text pairs spanning 30 speakers we compare against five strong voice-cloning text-to-speech baselines: our enhanced 5B model attains the highest speaker-encoder cosine similarity (SECS) across three independent verification networks (ECAPA-TDNN, WavLM-SV, Resemblyzer), statistically significantly outperforming every baseline. A side product of the architecture is that the audio path can be evaluated without the video path at inference time, yielding a ~30x speed-up over the full audio-video diffusion loop while preserving the voice-cloning behaviour.
Read the full abstract

All authors Ivan Mikheev, Viacheslav Vasilev, Anna Dmitrienko, Alexey Letunovskiy, Ivan Kirillov, Kirill Chernyshev, Denis Dimitrov

Saturday 15 August

Multimedia · Cross-listed from cs.RO

VTInstructor: Visual Trajectory Prompting for Navigation Instruction Generation in Continuous Environments

Haolin Yang, Yuxing Long, Zihan Yang et al. (+1 more)

Navigation instruction generation from ego-centric RGB video in continuous environments is an important yet challenging task for human-robot interaction and scalable dataset construction. Prior instruction generators assume discrete viewpoint graphs with panoramic observations, where trajectory structure is explicit; in continuous environments, however, the agent receives only a dense RGB stream, making trajectory cues difficult to recover. We propose VTInstructor, the first VLN instruction generation framework for continuous environments. Our key idea is to convert implicit trajectory geometry into explicit visual trajectory prompts: EDTC condenses long RGB trajectories into navigation-critical keyframes, VTP overlays path, turn, and goal cues onto these anchors, VTMod injects the resulting trajectory signals into the visual encoder, and VT-GRPO further calibrates this spatial injection during training, all without requiring a navigation graph, pre-built map, or scene reconstruction. On the challenging R2R-CE and RxR-CE Val Unseen benchmarks, VTInstructor sets a new state of the art across all standard NLG metrics, surpassing the strongest baseline by +0.357 CIDEr and +0.109 CIDEr, respectively. Beyond automatic metrics, VTInstructor-generated instructions raise a frozen follower's success rate to 63.3%, a +14.7 percentage-point gain over the best competing instruction source, and provide consistent data augmentation gains of +3 SR points on downstream navigation tasks.
Read the full abstract

All authors Haolin Yang, Yuxing Long, Zihan Yang, Hao Dong

Multimedia

RoE-FND: Synergizing LLMs with Experiential Learning for Effective and Generalizable Evidence-Based Fake News Detection

Yuzhou Yang, Qichao Ying, Sheng Li et al. (+3 more)

The proliferation of deceptive content in social networks necessitates robust Fake News Detection (FND) systems. Existing pipelines either train detectors on labeled data or leverage Large Language Models (LLMs) for their reasoning ability. However, current approaches remain either limited in generalizability or prone to over-commitment to persuasive yet flawed rationales, lacking systematic experience and mechanisms to expose subtle reasoning errors. We propose RoE-FND (\textbf{\underline{R}}eason \textbf{\underline{o}}n \textbf{\underline{E}}xperiences FND), an LLM-based framework that combines self-reflective experience building with deliberation through retrieved experiences for FND. RoE-FND builds an experience bank via reflective learning that compares an unconstrained analysis with a label-conditioned analysis using the ground-truth label as posterior supervision, then summarizes their critical divergence into reusable reasoning guidelines. During inference, RoE-FND generates two opposing deductions via a flipped pseudo-label provided as posterior, retrieves the most relevant experiences for resolving their key disagreement, and adjudicates the better-supported rationale as the final prediction. Experiments across five popular benchmarks, including text-only datasets, i.e., CHEF, Snopes, PolitiFact, and multimedia datasets, i.e., FakeTT, FakeSV, demonstrate that RoE-FND outperforms strong baselines without optimizing LLM parameters on dataset distributions, while exhibiting strong cross-dataset generalization.
Read the full abstract

All authors Yuzhou Yang, Qichao Ying, Sheng Li, Zhiyin Zhu, Zhenxing Qian, Xinpeng Zhang

Multimedia · Cross-listed from eess.SP

Flexible Deep Joint Source-Channel Coding: A Vibrotactile Example

Shuijie Li, Kemi Chen, Runjie Wang et al. (+2 more)

The increasing demand for real-time tactile communication in multimedia systems has exposed the limitations of existing Joint Source-Channel Coding (JSCC) techniques. While current JSCC models facilitate end-to-end optimization, they typically operate at fixed coding rates and require separate model instances for different rate settings. This results in significant storage overhead and limited adaptability to dynamic bandwidth conditions. To address these challenges, we propose the Flexible Deep Joint Source-Channel Coding (FD-JSCC) framework for vibrotactile signals, which supports flexible-rate transmission without the need for model switching. The FD-JSCC integrates a flexible-rate encoder-decoder enhanced with Hierarchical Gain Adaptation Module (HGAM) and Rate-Switchable Residual Module (RSRM), enabling bitrate-aware compression by selectively preserving salient vibrotactile features. Additionally, we introduce a Channel Feature Processing Module (CFPM), which leverages real-time SNR information to enhance robustness against channel noise and signal degradation. Trained on the IEEE 1918.1.1 vibrotactile dataset, FD-JSCC achieves reconstruction performance comparable to fixed-rate baselines (e.g., DeepSC-S), while reducing storage requirements by 61.1% when supporting four rates. These results underscore its potential for scalable, low-latency tactile communication in next-generation networks.
Read the full abstract

All authors Shuijie Li, Kemi Chen, Runjie Wang, Tiesong Zhao, Xiaoming Tao

Multimedia · Cross-listed from eess.SP

ParaJSCC: A Parameterized Framework for Reusable Multimodal Joint Source-Channel Coding

Kemi Chen, Mingkai Chen, Youjia Chen et al. (+3 more)

Multimodal signals, such as visual, audio, and tactile data, are increasingly maintained as persistent digital assets in immersive communication systems and digital twins. In these settings, the same multimodal content is repeatedly accessed by heterogeneous receivers with varying modality and bandwidth requirements. Existing compression and Joint Source-Channel Coding (JSCC) methods typically follow a per-request encoding paradigm, resulting in redundant computation and low efficiency during repeated access. To address this issue, we propose ParaJSCC, a multimodal JSCC framework designed for reusable representation serving. ParaJSCC converts each multimodal sample offline at the cloud/content server into a compact, quantized parameter package, which is then stored at the edge serving node for low-latency access. During serving, only the subset required by the current request is transmitted over the wireless channel, followed by lightweight decoding at the receiver. The framework employs a progressive shared-private parameterization to support modality-selective transmission and scalable reconstruction under varying bandwidth constraints. Experiments on multimodal datasets show that ParaJSCC significantly reduces online latency (e.g., from 17.18~ms to 4.34~ms for image-only requests and from 43.96~ms to 11.21~ms for full multimodal requests) and transmission rate (by 47.8%–51.2% for selective requests), while maintaining strong reconstruction quality under noisy channels.
Read the full abstract

All authors Kemi Chen, Mingkai Chen, Youjia Chen, Qian Liu, Wei Gao, Tiesong Zhao

Multimedia · Cross-listed from cs.CV

MetaReason: Precise Interleaved Multimodal Reasoning via Editing Meta Information for Solving Geometry Problems

Penghao Yin, Haomin Wang, Qihong Tang et al. (+3 more)

Although visual reasoning is crucial for solving complex geometry tasks, existing vision-language models rely heavily on text-only reasoning. Some recent methods introduce intermediate visual states to facilitate reasoning, but they are often hindered by inaccurate geometric representations and low rendering fidelity, ultimately leading to unreliable outputs. To address these limitations, we propose MetaReason, a framework for multimodal reasoning in plane geometry that leverages structured meta-information to enable accurate auxiliary-line construction. The framework first parses geometric images into meta-information, performs controllable edits with predefined tools to synthesize high-fidelity visual states, and then conducts reasoning based on these augmented views. To support this framework, we construct TutorGeo, a comprehensive dataset containing 17k image-to-meta conversion samples, 60k text-only reasoning traces, and 60k interleaved multimodal reasoning traces. Using this dataset, we combine supervised fine-tuning and reinforcement learning to develop robust multimodal reasoning capabilities. We also introduce ExamGeo, a benchmark derived from real-world examination problems that enables systematic evaluation across varying difficulty levels. Experimental results demonstrate that MetaReason significantly outperforms existing open-source models and achieves competitive performance against proprietary models.
Read the full abstract

All authors Penghao Yin, Haomin Wang, Qihong Tang, Xiaoye Qu, Hongjie Zhang, Xiao-Ping Zhang

Friday 14 August

Multimedia · Cross-listed from eess.IV

Personalized Digital Semantic Communication for Image Transmission with Vision-Language Models

Nan Li, Li Zhou, Haijun Wang et al. (+3 more)

Semantic communication (SC) enables bandwidth-efficient wireless image transmission, but most existing SC schemes are user-agnostic and ignore receiver-dependent semantics. To address this issue, we propose a personalized digital semantic communication (PDSC) framework that integrates a vision-language model (VLM)-based semantic encoder with a latent diffusion model (LDM)-based semantic decoder. Specifically, the semantic encoder extracts source-aware personalized semantic tokens from both the source image and the receiver's historical interactions. These tokens are vector-quantized into discrete semantic indices and further encoded into a compact fixed-length bitstream, enabling compatibility with digital transmission. At the receiver, the semantic decoder reconstructs a personalized image conditioned on the recovered semantic tokens. Furthermore, we formulate a capacity-constrained personalized semantic rate-distortion problem and introduce a semantic distortion metric that jointly characterizes source-semantic fidelity and user-preference alignment. Experiments show that PDSC achieves superior source-semantic consistency and personalization over state-of-the-art SC baselines, including CDDM and MoS, under bandwidth-limited wireless transmission.
Read the full abstract

All authors Nan Li, Li Zhou, Haijun Wang, Jun Xiong, Haitao Zhao, Jibo Wei

Multimedia

AlignFace: Human-Aligned Face Similarity Metric with Interpretable Concept Relations

Ying Huang, Wencan Zhang and Brian Y. Lim

Computer vision models for generated facial content, such as face editing and privacy protection, increasingly affect people, requiring similarity metrics that serve as faithful proxies for human perception. While perceptual evaluation has progressed from signal-based heuristics to representation-based metrics, current approaches are limited to behavioral modeling without cognitive alignment. They rely on implicit and spurious relations while assuming a universal observer, failing to account for inherent variations across diverse human populations. This leads to inaccurate evaluative models of stakeholders and misleading guidance for generative model debugging. Rather than treating perception as a black box, we leverage scientific findings from cognitive psychology of human face similarity perception: dependence on facial featural and configural attributes, nonlinear psychophysical response scaling, and own-group biases. We introduce the FACETS dataset and propose AlignFace, an interpretable, human-aligned, face similarity metric that encodes these cognitive principles through ante-hoc modeling. It employs visual-language modeling (VLM) to encode paired face images and text-based attributes, gated cross-attention (CA) to extract attribute-specific facial difference representations, concept bottleneck modeling (CBM) to constrain reasoning via interpretable face attributes, and neural generalized additive model (GAM) to model their nonlinear influence. Experiments found AlignFace significantly improves alignment with human subpopulation perceptions compared to baseline metrics, including recent domain-free learned perceptual metrics. By bridging learned representations and human cognitive processes, this work enables more transparent and aligned perceptual evaluation metrics for face images.
Read the full abstract

Multimedia · Cross-listed from cs.SD

H2H Music Improv: A Communication Model and Audio-Visual Dataset for Music Improvisation

Aleksandra Teng Ma, Anthony Cammarota, Jiayi Wang et al. (+4 more)

Current real-time AI improvisation systems lack the communication awareness human musicians rely on: rather than treating communication as a foundational algorithm design concern, most systems layer interaction strategies post-hoc onto generative algorithms through explicit controls and predefined modes. This gap persists in part because no formalized, machine-readable communication model with musicians exists. To address this, we study how expert musicians communicate in free (non-idiomatic) improvisation, unconstrained by prior discussion or agreement. Through a collaborative co-design process with expert improvisers, we derive a communication model that (1) captures how free improvisers negotiate musical ideas and enter stable musical spaces, and (2) is formalized as a machine-readable annotation scheme. We further present the H2H (Human-to-Human) Music Improvisation dataset: six hours of audio-visual expert duo improvisations with clean per-player stems and per-player annotations of both their own intentions and their perception of their partner's intentions. To our knowledge, this is the first such dataset for free improvisation. Together, the communication model and the dataset offer a new lens and resource for studying musician communication and may in future inform the design of AI musical partners that communicate by design.
Read the full abstract

All authors Aleksandra Teng Ma, Anthony Cammarota, Jiayi Wang, Alexandria Smith, Cheng-Zhi Anna Huang, Jeffrey Albert, Alexander Lerch

Thursday 13 August

Multimedia · Cross-listed from cs.CV

NARU: A Benchmark for NARrative Evolution and Cultural Nuance Understanding in Japanese Extreme Long Video

Yuheng Huang, Jianlang Chen, Jiayang Song et al. (+6 more)

Long-form video understanding encompasses tasks that go beyond retrieving isolated events, including tracking an evolving narrative and interpreting social meaning that may remain implicit. However, existing benchmarks rarely evaluate these capabilities jointly, particularly in high-context, non-English media. To address this gap, we introduce NARU, a benchmark designed to evaluate Narrative evolution and Reasoning on cultural Understanding in Japanese long-form video. NARU consists of 1,481 questions grounded in 155 videos totaling 146.8 hours, spanning four narrative and five cultural dimensions. To construct the benchmark at this scale, we propose a hierarchical memory-based annotation pipeline that transforms raw video into structured event, narrative, and cultural annotations, then generates questions via task-oriented synthesis and iterative shortcut removal. The construction process includes two native-speaker verification stages involving 68 annotators. Evaluations across eight model configurations reveal substantial limitations in both long-range narrative integration and culturally grounded reasoning. By exposing these persistent gaps, NARU offers a systematic testing ground for developing MLLMs capable of reliably interpreting long-form, high-context video.
Read the full abstract

All authors Yuheng Huang, Jianlang Chen, Jiayang Song, Hua Qi, Aza Kai, Vincent Markert, Edison Marrese-Taylor, Jianjun Zhao, Lei Ma

Multimedia · Cross-listed from cs.CV

Beyond Visual Evidence: Revealing and Mitigating Relational Privacy Leakage in Document MLLMs

Beining Xu, Hairui Wang, Jiaxin Wang et al. (+2 more)

While the privacy risks of multimodal large language models (MLLMs) have drawn significant attention, the unique vulnerabilities of domain-specific MLLMs remain largely underexplored. Focusing on document understanding MLLMs for identity document processing, this paper investigates the privacy issues inherent in Key Information Extraction (KIE) tasks. We reveal that when input images lack sufficient visual evidence, these models often rely on memorized field relations from training data to infer missing content, thereby leaking multiple correlated fields containing sensitive personal information. To mitigate this risk, we make three key contributions.First, we propose the Dynamic Relational Unlearning Framework (DRUF) which comprises a Relational Decoupling Unlearning (RDU) module and a dynamic set update mechanism. It suppresses the leakage of high-risk field pairs while preserving KIE performance.Second, we introduce DocPrivacyBench, a novel benchmark to systematically evaluate a model's susceptibility to privacy leakage under conditions of absent or minimal visual evidence.Third, we evaluate three MLLMs and six unlearning methods using this benchmark, assessing both post-unlearning leakage suppression and utility preservation.Our results demonstrate that existing MLLMs consistently exhibit privacy leakage when visual evidence is scarce, particularly on noisier datasets. In contrast, DRUF outperforms the strongest baseline by improving leakage suppression by 4.8 percentage points, effectively mitigating privacy risks while maintaining robust document information extraction performance.
Read the full abstract

All authors Beining Xu, Hairui Wang, Jiaxin Wang, Changsheng Chen, Anirban Chakraborty