Audio-Visual Understanding

Recent advances in reasoning models have shown remarkable progress in text-based domains, but transferring those capabilities to multimodal settings, e.g., to allow reasoning over audio-visual data, still remains a challenge, in part because of the limited availability of high-quality reasoning data in targeted multimodal combinations. To address this problem, we introduce AVRT, a novel framework that generates high-quality audio-visual reasoning traces from single-modality teacher models. We generate independent vision- and audio-reasoning traces via models specialized to reason over their respective modalities and merge the resulting traces with an LLM merger model. The resulting multimodal traces are used in a supervised fine-tuning (SFT) cold start to adapt the target model to audio-visual reasoning traces first, before training it in a second reinforcement learning stage on larger-scale data. Evaluated on seven audio-visual and audio benchmarks, our 3B and 7B parameter models achieve state-of-the-art results among models of comparable size including OmniBench and DailyOmni for audio-visual and MMAR for audio-only reasoning, showing that cross-modal training also transfers to single-modality tasks and establishing a new training pipeline for multimodal reasoning models.

In this work, we present Qwen3.5-Omni, the latest advancement in the Qwen-Omni model family. Representing a significant evolution over its predecessor, Qwen3.5-Omni scales to hundreds of billions of parameters and supports a 256k context length. By leveraging a massive dataset comprising heterogeneous text-vision pairs and over 100 million hours of audio-visual content, the model demonstrates robust omni-modality capabilities. Qwen3.5-Omni-plus achieves SOTA results across 215 audio and audio-visual understanding, reasoning, and interaction subtasks and benchmarks, surpassing Gemini-3.1 Pro in key audio tasks and matching it in comprehensive audio-visual understanding. Architecturally, Qwen3.5-Omni employs a Hybrid Attention Mixture-of-Experts (MoE) framework for both Thinker and Talker, enabling efficient long-sequence inference. The model facilitates sophisticated interaction, supporting over 10 hours of audio understanding and 400 seconds of 720P video (at 1 FPS). To address the inherent instability and unnaturalness in streaming speech synthesis, often caused by encoding efficiency discrepancies between text and speech tokenizers, we introduce ARIA. ARIA dynamically aligns text and speech units, significantly enhancing the stability and prosody of conversational speech with minimal latency impact. Furthermore, Qwen3.5-Omni expands linguistic boundaries, supporting multilingual understanding and speech generation across 10 languages with human-like emotional nuance. Finally, Qwen3.5-Omni exhibits superior audio-visual grounding capabilities, generating script-level structured captions with precise temporal synchronization and automated scene segmentation. Remarkably, we observed the emergence of a new capability in omnimodal models: directly performing coding based on audio-visual instructions, which we call Audio-Visual Vibe Coding.

Audio tokenization has emerged as a critical component in end-to-end audio language models, enabling efficient discrete representation learning for both audio understanding and generation tasks. However, existing audio tokenizers face fundamental limitations in understanding tasks due to single-modality constraints, particularly when audio signals contain ambiguous or incomplete information. While incorporating additional modality information can significantly enhance audio understanding, current multimodal fusion approaches invariably degrade reconstruction quality. This degradation is unacceptable for end-to-end audio systems that require high-fidelity audio generation capabilities. In this work, we investigate the root causes of reconstruction quality degradation in video-enhanced audio tokenization and present three key findings. First, the location of fusion within the tokenizer architecture is crucial for preserving reconstruction quality. Second, we show that contrastive learning, though effective in continuous representation fusion, is unsuitable for discrete tokenizers as it fails to enhance downstream task performance. Third, while feature-dimension fusion approaches achieve moderate success, we discover that fusing along the temporal axis -- guided by the concept of distinctive features -- yields significantly better results. Building on these insights, we introduce the Timing-Aware Pre-Quantization Fusion for Video-Enhanced Audio Tokenization, the first approach to successfully integrate visual information into audio tokenizer architectures while preserving reconstruction fidelity. Our approach not only maintains high-fidelity reconstruction but also achieves superior performance on downstream understanding tasks compared with audio-only tokenizers and established multimodal fusion baselines.

To extend the reinforcement learning post-training paradigm to omni-modal models for concurrently bolstering video-audio understanding and collaborative reasoning, we propose OmniJigsaw, a generic self-supervised framework built upon a temporal reordering proxy task. Centered on the chronological reconstruction of shuffled audio-visual clips, this paradigm strategically orchestrates visual and auditory signals to compel cross-modal integration through three distinct strategies: Joint Modality Integration, Sample-level Modality Selection, and Clip-level Modality Masking. Recognizing that the efficacy of such proxy tasks is fundamentally tied to puzzle quality, we design a two-stage coarse-to-fine data filtering pipeline, which facilitates the efficient adaptation of OmniJigsaw to massive unannotated omni-modal data. Our analysis reveals a ``bi-modal shortcut phenomenon'' in joint modality integration and demonstrates that fine-grained clip-level modality masking mitigates this issue while outperforming sample-level modality selection. Extensive evaluations on 15 benchmarks show substantial gains in video, audio, and collaborative reasoning, validating OmniJigsaw as a scalable paradigm for self-supervised omni-modal learning.

Sound source localization task aims to identify the locations of sound-emitting objects by leveraging correlations between audio and visual modalities. Most existing SSL methods rely on contrastive learning-based feature matching, but lack explicit reasoning and verification, limiting their effectiveness in complex acoustic scenes. Inspired by human meta-cognitive processes, we propose a training-free SSL framework that exploits the intrinsic reasoning capabilities of Multimodal Large Language Models (MLLMs). Our Generation-Analysis-Refinement (GAR) pipeline consists of three stages: Generation produces initial bounding boxes and audio classifications; Analysis quantifies Audio-Visual Consistency via open-set role tagging and anchor voting; and Refinement applies adaptive gating to prevent unnecessary adjustments. Extensive experiments on single-source and multi-source benchmarks demonstrate competitive performance. The source code is available at https://github.com/VisualAIKHU/GAR-SSL.

As audio-visual multi-modal large language models (MLLMs) are increasingly deployed in safety-critical applications, understanding their cross-modal vulnerabilities is essential. To this end, we introduce Multi-Modal Typography, a systematic study examining how perturbations added to speech and on-screen text impact audio-visual reasoning of closed- and open-source models. For instance, on the WorldSense benchmark, the attack success rate (ASR) against Gemini 3.7 Flash increased from 9.53% to 43.41%, highlighting its vulnerability. We analyze the cross-modal effects of unimodal and multimodal attacks and evaluate how specific properties of the attacks impact the overall multi-modal reasoning. By applying fine-tuning as a mitigation, we lower the ASR of Qwen2.5-Omni-7B on WorldSense from $60.6\%$ to $52.0\%$. Our findings across multi-modal reasoning tasks, content moderation, and offline robot task selection benchmarks establish multi-modal typography as a critical and underexplored attack strategy in multi-modal reasoning. Code and data will be made publicly available.

Audio-Visual Large Language Models (AVLLMs) are emerging as unified interfaces to multimodal perception. We present the first mechanistic interpretability study of AVLLMs, analyzing how audio and visual features evolve and fuse through different layers of an AVLLM to produce the final text outputs. We find that although AVLLMs encode rich audio semantics at intermediate layers, these capabilities largely fail to surface in the final text generation when audio conflicts with vision. Probing analyses show that useful latent audio information is present, but deeper fusion layers disproportionately privilege visual representations that tend to suppress audio cues. We further trace this imbalance to training: the AVLLM's audio behavior strongly matches its vision-language base model, indicating limited additional alignment to audio supervision. Our findings reveal a fundamental modality bias in AVLLMs and provide new mechanistic insights into how multimodal LLMs integrate audio and vision.

Automatic violence detection from video is challenging because violent interactions may be distant, occluded, or only partially visible. Audio can provide complementary evidence for violent events that are difficult to recognize from visual information alone. However, audio itself may be absent, dubbed, or dominated by environmental noise, making the central challenge not whether to incorporate audio but how to adapt reliance on it according to the visual scene. We introduce AViS-Mamba, an audiovisual Mamba-based architecture in which the visual stream directly governs the behavior of the audio stream. At each layer of the audio encoder, a compact visual representation produces a modulation vector that conditions the encoder's internal temporal operators together with a routing gate that regulates the strength of this visual intervention. Rather than fusing or reweighting features after they have been extracted, visual context directly shapes the temporal dynamics of the audio encoder. We further propose Adaptive AV-InfoNCE, a contrastive objective that learns to balance the audio-to-video and video-to-audio alignment directions rather than weighting them uniformly. On the audio-valid NTU-CCTV and DVD benchmarks, AViS-Mamba establishes state-of-the-art results, attaining 88.59% and 75.74% accuracy. We demonstrate that adaptive visual conditioning consistently outperforms fixed routing and improves performance under degraded and missing-audio conditions. Layer-wise analysis further reveals that the model adapts the audio stream selectively across network depth rather than applying a single global routing policy.

We introduce Omni-MMSI, a new task that requires comprehensive social interaction understanding from raw audio, vision, and speech input. The task involves perceiving identity-attributed social cues (e.g., who is speaking what) and reasoning about the social interaction (e.g., whom the speaker refers to). This task is essential for developing AI assistants that can perceive and respond to human interactions. Unlike prior studies that operate on oracle-preprocessed social cues, Omni-MMSI reflects realistic scenarios where AI assistants must perceive and reason from raw data. However, existing pipelines and multi-modal LLMs perform poorly on Omni-MMSI because they lack reliable identity attribution capabilities, which leads to inaccurate social interaction understanding. To address this challenge, we propose Omni-MMSI-R, a reference-guided pipeline that produces identity-attributed social cues with tools and conducts chain-of-thought social reasoning. To facilitate this pipeline, we construct participant-level reference pairs and curate reasoning annotations on top of the existing datasets. Experiments demonstrate that Omni-MMSI-R outperforms advanced LLMs and counterparts on Omni-MMSI. Project page: https://sampson-lee.github.io/omni-mmsi-project-page.

Binaural audio delivers spatial cues essential for immersion, yet most consumer videos are monaural due to capture constraints. We introduce SIREN, a visually guided mono to binaural framework that explicitly predicts left and right channels. A ViT-based encoder learns dual-head self-attention to produce a shared scene map and end-to-end L/R attention, replacing hand-crafted masks. A soft, annealed spatial prior gently biases early L/R grounding, and a two-stage, confidence-weighted waveform-domain fusion (guided by mono reconstruction and interaural phase consistency) suppresses crosstalk when aggregating multi-crop and overlapping windows. Evaluated on FAIR-Play and MUSIC-Stereo, SIREN yields consistent gains on time-frequency and phase-sensitive metrics with competitive SNR. The design is modular and generic, requires no task-specific annotations, and integrates with standard audio-visual pipelines.

Every year, 10 million pets enter shelters, separated from their families. Despite desperate searches by both guardians and lost animals, 70% never reunite, not because matches do not exist, but because current systems look only at appearance, while animals recognize each other through sound. We ask, why does computer vision treat vocalizing species as silent visual objects? Drawing on five decades of cognitive science showing that animals perceive quantity approximately and communicate identity acoustically, we present the first multimodal reunification system integrating visual and acoustic biometrics. Our species-adaptive architecture processes vocalizations from 10Hz elephant rumbles to 4kHz puppy whines, paired with probabilistic visual matching that tolerates stress-induced appearance changes. This work demonstrates that AI grounded in biological communication principles can serve vulnerable populations that lack human language.

Large-scale pre-trained image-text models exhibit robust multimodal representations, yet applying the Contrastive Language-Image Pre-training (CLIP) model to audio-visual localization remains challenging. Replacing the classification token ([CLS]) with an audio-embedded token ([V_A]) struggles to capture semantic cues, and the prompt "a photo of a [V_A]" fails to establish meaningful connections between audio embeddings and context tokens. To address these issues, we propose Sound-aware Prompt Learning (SOUPLE), which replaces fixed prompts with learnable context tokens. These tokens incorporate visual features to generate conditional context for a mask decoder, effectively bridging semantic correspondence between audio and visual inputs. Experiments on VGGSound, SoundNet, and AVSBench demonstrate that SOUPLE improves localization and segmentation performance.

While Omni-modal Large Language Models have made strides in joint sensory processing, they fundamentally struggle with a cornerstone of human interaction: deciphering complex, multi-person conversational dynamics to accurately answer Who said what and when.'' Current models suffer from an illusion of competence'' -- they exploit visual biases in conventional benchmarks to bypass genuine cross-modal alignment, while relying on sparse, low-frame-rate visual sampling that destroys crucial high-frequency dynamics like lip movements. To shatter this illusion, we introduce Visual-Registered Speaker Diarization and Recognition (VR-SDR) and the HumanOmni-Speaker Benchmark. By strictly eliminating visual shortcuts, this rigorous paradigm demands true end-to-end spatio-temporal identity binding using only natural language queries. To overcome the underlying architectural perception gap, we propose HumanOmni-Speaker, powered by a Visual Delta Encoder. By sampling raw video at 25 fps and explicitly compressing inter-frame motion residuals into just 6 tokens per frame, it captures fine-grained visemes and speaker trajectories without triggering a catastrophic token explosion. Ultimately, HumanOmni-Speaker demonstrates strong multimodal synergy, natively enabling end-to-end lip-reading and high-precision spatial localization without intrusive cropping, and achieving superior performance across a wide spectrum of speaker-centric tasks.

Recent advancements in omnimodal large language models (OmniLLMs) have significantly improved the comprehension of audio and video inputs. However, current evaluations primarily focus on short audio and video clips ranging from 10 seconds to 5 minutes, failing to reflect the demands of real-world applications, where videos typically run for tens of minutes. To address this critical gap, we introduce LVOmniBench, a new benchmark designed specifically for the cross-modal comprehension of long-form audio and video. This dataset comprises high-quality videos sourced from open platforms that feature rich audio-visual dynamics. Through rigorous manual selection and annotation, LVOmniBench comprises 275 videos, ranging in duration from 10 to 90 minutes, and 1,014 question-answer (QA) pairs. LVOmniBench aims to rigorously evaluate the capabilities of OmniLLMs across domains, including long-term memory, temporal localization, fine-grained understanding, and multimodal perception. Our extensive evaluation reveals that current OmniLLMs encounter significant challenges when processing extended audio-visual inputs. Open-source models generally achieve accuracies below 35%, whereas the Gemini 3 Pro reaches a peak accuracy of approximately 65%. We anticipate that this dataset, along with our empirical findings, will stimulate further research and the development of advanced models capable of resolving complex cross-modal understanding problems within long-form audio-visual contexts.

Evaluating omni-modal large language models (OLMs) in multi-party dialogue requires more than answer correctness on pre-segmented inputs. We introduce SocialOmni, an offline diagnostic benchmark that separates three turn-level decisions: identifying who is speaking, deciding when a designated participant should enter at an annotated query time, and determining how that participant should continue the dialogue. SocialOmni contains 2,000 perception items and a quality-controlled core split of 200 interaction-generation items, including naturally occurring speaker-visibility mismatches. Each item is independently checked by three human annotators. Evaluated systems receive only query-time-bounded multimodal evidence, while manually verified reference continuations are reserved for response judging. A complete three-judge ensemble scores every eligible response, with leave-one-judge-out and family-sensitivity audits. Across 11 OLMs, rankings vary substantially by axis, and coverage-adjusted scores reveal when high conditional response quality depends on selective turn entry. The protocol does not measure persistent streaming state or wall-clock latency.

Omnimodal large language models (OmniLLMs) jointly process audio and visual streams, but the resulting long multimodal token sequences make inference prohibitively expensive. Existing compression methods typically rely on fixed window partitioning and attention-based pruning, which overlook the piecewise semantic structure of audio-visual signals and become fragile under aggressive token reduction. We propose Dynamic Audio-driven Semantic cHunking (DASH), a training-free framework that aligns token compression with semantic structure. DASH treats audio embeddings as a semantic anchor and detects boundary candidates via cosine-similarity discontinuities, inducing dynamic, variable-length segments that approximate the underlying piecewise-coherent organization of the sequence. These boundaries are projected onto video tokens as a soft temporally co-registered segmentation prior. Within each segment, token retention is determined by a tri-signal importance estimator that fuses structural boundary cues, representational distinctiveness, and attention-based salience, mitigating the sparsity bias of attention-only selection. This structure-aware allocation preserves transition-critical tokens while reducing redundant regions. Extensive experiments on AVUT, VideoMME, and WorldSense demonstrate that DASH maintains competitive or superior accuracy while achieving higher compression ratios compared to prior methods. Code is available at: https://github.com/laychou666/DASH.

The ability to capture and segment sounding objects in dynamic visual scenes is crucial for the development of Audio-Visual Segmentation (AVS) tasks. While significant progress has been made in this area, the interaction between audio and visual modalities still requires further exploration. In this work, we aim to answer the following questions: How can a model effectively suppress audio noise while enhancing relevant audio information? How can we achieve discriminative interaction between the audio and visual modalities? To this end, we propose SDAVS, equipped with the Selective Noise-Resilient Processor (SNRP) module and the Discriminative Audio-Visual Mutual Fusion (DAMF) strategy. The proposed SNRP mitigates audio noise interference by selectively emphasizing relevant auditory cues, while DAMF ensures more consistent audio-visual representations. Experimental results demonstrate that our proposed method achieves state-of-the-art performance on benchmark AVS datasets, especially in multi-source and complex scenes. The code and model are available at https://github.com/happylife-pk/SDAVS.

Multimodal Large Language Models (MLLMs) have shown strong performance in visual and audio understanding when evaluated in isolation. However, their ability to jointly reason over omni-modal (visual, audio, and textual) signals in long and complex videos remains largely unexplored. We introduce MMOU, a new benchmark designed to systematically evaluate multimodal understanding and reasoning under these challenging, real-world conditions. MMOU consists of 20,000 carefully curated questions paired with 11877 web-collected videos of varying length, spanning diverse domains and exhibiting rich, tightly coupled audio-visual content. The benchmark covers 13 fundamental skill categories, all of which require integrating evidence across modalities and time. All questions are manually annotated across multiple turns by professional annotators, ensuring high quality and reasoning fidelity. We evaluate 20+ state-of-the-art open-source and proprietary multimodal models on MMOU. The results expose substantial performance gaps: the best closed-source model achieves only 64.2% accuracy, while the strongest open-source model reaches just 46.8%. Our results highlight the challenges of long-form omni-modal understanding, revealing that current models frequently fail to apply even fundamental skills in long videos. Through detailed analysis, we further identify systematic failure modes and provide insights into where and why current models break.

Audio-Visual Learning (AVL) is one fundamental task of multi-modality learning and embodied intelligence, displaying the vital role in scene understanding and interaction. However, previous researchers mostly focus on exploring downstream tasks from a coarse-grained perspective (e.g., audio-visual correspondence, sound source localization, and audio-visual event localization). Considering providing more specific scene perception details, we newly define a fine-grained Audio-Visual Learning task, termed Region-Aware Sound Source Understanding (RA-SSU), which aims to achieve region-aware, frame-level, and high-quality sound source understanding. To support this goal, we innovatively construct two corresponding datasets, i.e. fine-grained Music (f-Music) and fine-grained Lifescene (f-Lifescene), each containing annotated sound source masks and frame-by-frame textual descriptions. The f-Music dataset includes 3,976 samples across 22 scene types related to specific application scenarios, focusing on music scenes with complex instrument mixing. The f-Lifescene dataset contains 6,156 samples across 61 types representing diverse sounding objects in life scenarios. Moreover, we propose SSUFormer, a Sound-Source Understanding TransFormer benchmark that facilitates both the sound source segmentation and sound region description with a multi-modal input and multi-modal output architecture. Specifically, we design two modules for this framework, Mask Collaboration Module (MCM) and Mixture of Hierarchical-prompted Experts (MoHE), to respectively enhance the accuracy and enrich the elaboration of the sound source description. Extensive experiments are conducted on our two datasets to verify the feasibility of the task, evaluate the availability of the datasets, and demonstrate the superiority of the SSUFormer, which achieves SOTA performance on the Sound Source Understanding benchmark.

Audio-Visual Segmentation (AVS) aims to produce pixel-level masks of sound producing objects in videos, by jointly learning from audio and visual signals. However, real-world environments are inherently dynamic, causing audio and visual distributions to evolve over time, which challenge existing AVS systems that assume static training settings. To address this gap, we introduce the first exemplar-free continual learning benchmark for Audio-Visual Segmentation, comprising four learning protocols across single-source and multi-source AVS datasets. We further propose a strong baseline, ATLAS, which uses audio-guided pre-fusion conditioning to modulate visual feature channels via projected audio context before cross-modal attention. Finally, we mitigate catastrophic forgetting by introducing Low-Rank Anchoring (LRA), which stabilizes adapted weights based on loss sensitivity. Extensive experiments demonstrate competitive performance across diverse continual scenarios, establishing a foundation for lifelong audio-visual perception. Code is available at${}^{*}$\footnote{Paper under review} - \hyperlink{https://gitlab.com/viper-purdue/atlas}{https://gitlab.com/viper-purdue/atlas} \keywords{Continual Learning \and Audio-Visual Segmentation \and Multi-Modal Learning}

When humans perceive the world, they naturally integrate multiple audio-visual tasks within dynamic, real-world scenes. However, current works such as event localization, parsing, segmentation and question answering are mostly explored individually, making it challenging to comprehensively understand complex audio-visual scenes and explore inter-task relationships. Hence, we propose AV-Unified, a unified framework that enables joint learning across a wide range of audio-visual scene understanding tasks. AV-Unified standardizes the diverse input-output formats of each task and incorporates a multi-scale spatiotemporal perception network to effectively capture audio-visual associations. Specifically, we unify the inputs and outputs of all supported tasks by converting them into sequences of discrete tokens, establishing a shared representation that allows a single architecture to be trained jointly across heterogeneous varied datasets. Considering the varying temporal granularity of audio-visual events, a multi-scale temporal perception module is designed to capture key cues. Meanwhile, to overcome the lack of auditory supervision in the visual domain, we design a cross-modal guidance-based spatial perception module that models spatial audio-visual associations. Furthermore, task-specific text prompts are employed to enhance the model's adaptability and task-awareness. Extensive experiments on benchmark datasets (e.g., AVE, LLP, MUSIC-AVQA, VGG-SS and AVS) demonstrate the effectiveness of AV-Unified across temporal, spatial, and spatiotemporal tasks.

While recent advances in Multimodal Large Language Models (MLLMs) have improved image-based localization, precise global geolocation remains a formidable challenge due to the inherent ambiguity of visual landscapes and the largely untapped potential of auditory cues. In this paper, we introduce Audiovisual Geolocation, a framework designed to resolve geographic ambiguity through interpretable perception and reasoning. We present AVG, a high-quality global-scale video benchmark for geolocation, comprising 20,000 curated clips across 1,000 distinct locations. To address the complexity of audiovisual geolocation, we propose a three-stage framework: (1) a Perception stage that utilizes a mixture-autoregressive sparse autoencoder to decompose noisy audio into semantically grounded "acoustic atoms"; (2) a Multimodal Reasoning stage that employs an MLLM finetuned via Group Relative Policy Optimization (GRPO) to synthesize these atoms with visual features; and (3) a Precision Prediction stage using Riemannian Flow Matching on the $S^2$ manifold. Our experiments demonstrate that our framework significantly outperforms unimodal baselines. These results entail that interpretable perception of the soundscape provides a critical, orthogonal signal that, when coupled with multimodal reasoning, enables high-precision global localization.

Developing Audio-Visual Large Language Models (AV-LLMs) for unified scene understanding is pivotal in multimodal intelligence. While instruction tuning enables pre-trained models with multi-task abilities, we observe that conventional multi-task unification methods often suffer from severe negative transfer, where nearly 55% of tasks degrade compared to single-task training. We attribute this phenomenon to audio-visual task heterogeneity, characterized by disparate task granularity and divergent capability demands, which lead to negative interference under joint training. To tackle this, we present Crab$^{+}$, a scalable and unified audio-visual scene understanding model that addresses task heterogeneity through explicit cooperation from both data and model perspectives. On the data side, we introduce AV-UIE v2, a comprehensive Audio-Visual Unified Instruction-tuning dataset with Explicit reasoning processes. It contains approximately 222K samples spanning 17 datasets and 7 tasks, enabling the model to capture cross-task relationships at different levels of granularity. On the model side, we design a unified interface to align heterogeneous task formulations, and propose Interaction-aware LoRA (I-LoRA), which explicitly models inter-task relationships via dynamic routing to coordinate distinct audio-visual interaction patterns, mitigating parameter interference. Extensive experiments show Crab$^{+}$ covers broader tasks than existing unified models while outperforming specialized models on various benchmarks. We successfully reverse the negative transfer trend, achieving positive transfer where multi-task learning surpasses single-task baselines in nearly 88% of tasks. These results hold across diverse AV-LLM paradigms and are validated through in-depth visualization, positioning Crab$^{+}$ as a robust step towards holistic audio-visual scene understanding.

Audio-visual continual test-time adaptation involves continually adapting a source audio-visual model at test-time, to unlabeled non-stationary domains, where either or both modalities can be distributionally shifted, which hampers online cross-modal learning and eventually leads to poor accuracy. While previous works have tackled this problem, we find that SOTA methods suffer from catastrophic forgetting where the model's performance drops well below even the source model due to continual parameter updates at test-time. In this work, we first show that adapting only the modality fusion layer to a target domain not only improves performance on that domain but can also enhance performance on subsequent domains. Based on this strong cross-task transferability of the fusion layer's parameters, we propose a method, $\texttt{AVReCAP}$, that improves test-time performance of the models without access to any source data. Our approach works by using a selective parameter retrieval mechanism that dynamically retrieves the best fusion layer parameters from a buffer using only a small batch of test data. These parameters are then integrated into the model, adapted to the current test distribution, and saved back for future use. Extensive experiments on benchmark datasets involving unimodal and bimodal corruptions show our proposed $\texttt{AVReCAP}$ significantly outperforms existing methods while minimizing catastrophic forgetting.

Large Audio Language Models struggle to disentangle overlapping events in complex acoustic scenes, yielding temporally inconsistent captions and frequent hallucinations. We introduce Timestamped Audio Captioner (TAC), a model that produces temporally grounded audio descriptions at varying degrees of detail and resolution. TAC is trained with a synthetic data pipeline that constructs challenging and dynamic mixtures from real-world audio sources, enabling robust learning under realistic polyphonic conditions. Across event detection and dense captioning, TAC outperforms all competing methods, with a low hallucination rate and accurate temporal grounding. We also introduce TAC-V, an audio-visual pipeline to generate semantically rich audio-visual descriptions. We then show that TAC and TAC-V serves as a "semantic bridge" for a text-only reasoner: a simple TAC$\rightarrow$LLM and TAC-V$\rightarrow$LLM cascade achieves state-of-the-art scores on benchmarks for both audio (MMAU-Pro, MMSU, MMAR) and audio-visual (DailyOmni, VideoHolmes) understanding and reasoning respectively.

This paper proposes Omni Dense Captioning, a novel task designed to generate continuous, fine-grained, and structured audio-visual narratives with explicit timestamps. To ensure dense semantic coverage, we introduce a six-dimensional structural schema to create "script-like" captions, enabling readers to vividly imagine the video content scene by scene, akin to a cinematographic screenplay. To facilitate research, we construct OmniDCBench, a high-quality, human-annotated benchmark, and propose SodaM, a unified metric that evaluates time-aware detailed descriptions while mitigating scene boundary ambiguity. Furthermore, we construct a training dataset, TimeChatCap-42K, and present TimeChat-Captioner-7B, a strong baseline trained via SFT and GRPO with task-specific rewards. Extensive experiments demonstrate that TimeChat-Captioner-7B achieves state-of-the-art performance, surpassing Gemini-2.5-Pro, while its generated dense descriptions significantly boost downstream capabilities in audio-visual reasoning (DailyOmni and WorldSense) and temporal grounding (Charades-STA). All datasets, models, and code are available at https://github.com/yaolinli/TimeChat-Captioner.

Multimodal Large Language Models (MLLMs) are a major focus of recent AI research. However, most prior work focuses on static image understanding, while their ability to process sequential audio-video data remains underexplored. This gap highlights the need for a high-quality benchmark to systematically evaluate MLLM performance in a real-world setting. We introduce SONIC-O1, a comprehensive, fully human-verified benchmark of 60 hours (231 clips) spanning 13 real-world conversational domains with 4,958 annotations and demographic metadata. SONIC-O1 evaluates three capabilities: open-ended summarization, multiple-choice question (MCQ) answering, and temporal localization with supporting rationales (reasoning). Across closed- and open-source models, we find that the MCQ accuracy shows the smallest gap between model families, but the best closed-source model outperforms the best open-source model by 22.6% on temporal localization. We further observe accuracy gaps of up to 21.4% on temporal localization across demographic groups, indicating persistent disparities in model behaviour. SONIC-O1 provides an open evaluation suite for temporally grounded and demographically robust multimodal understanding. SONIC-O1 is publicly available for research: Project page (https://vectorinstitute.github.io/sonic-o1/), Dataset (https://huggingface.co/datasets/vector-institute/sonic-o1), GitHub (https://github.com/vectorinstitute/sonic-o1), Leaderboard (https://huggingface.co/spaces/vector-institute/sonic-o1-leaderboard).

Internet audio-visual clips convey meaning through time-varying sound and motion, which extend beyond what text alone can represent. To examine whether AI models can understand such signals in human cultural contexts, we introduce AVMeme Exam, a human-curated benchmark of over one thousand iconic Internet sounds and videos spanning speech, songs, music, and sound effects. Each meme is paired with a unique Q&A assessing levels of understanding from surface content to context and emotion to usage and world knowledge, along with metadata such as original year, transcript, summary, and sensitivity. We systematically evaluate state-of-the-art multimodal large language models (MLLMs) alongside human participants using this benchmark. Our results reveal a consistent limitation: current models perform poorly on textless music and sound effects, and struggle to think in context and in culture compared to surface content. These findings highlight a key gap in human-aligned multimodal intelligence and call for models that can perceive contextually and culturally beyond the surface of what they hear and see. Project page: avmemeexam.github.io/public

Audio-Visual Segmentation (AVS) aims to localize sound-producing objects at the pixel level by integrating auditory and visual cues. However, existing methods often struggle with multi-source entanglement and audio-visual misalignment, leading to a dominance bias toward acoustically or visually salient objects (i.e., louder or larger ones) at the expense of subtler or co-occurring sources. To address these challenges, we propose DDAVS: Delayed Bidirectional Alignment via Disentangled Audio Semantics for Audio-Visual Segmentation. To mitigate multi-source entanglement, DDAVS employs learnable queries to extract audio semantics and anchor them within a structured semantic space derived from an audio prototype memory bank. This process is further optimized through contrastive learning to enhance discriminability and robustness. To alleviate audio-visual misalignment, DDAVS introduces dual cross attention with delayed modality interaction, improving the robustness of multimodal alignment. Extensive experiments on the AVS-Objects and VPO benchmarks demonstrate that DDAVS achieves state-of-the-art performance across single-source, multi-source, and multi-class multi-instance scenarios. These results validate the effectiveness and generalization ability of our framework under challenging real-world audio-visual segmentation conditions. Project page: https://trilarflagz.github.io/DDAVS-page/

Recently, significant progress has been made in multi-modal continual learning, aiming to learn new tasks sequentially in multi-modal settings while preserving performance on previously learned ones. However, existing methods mainly focus on coarse-grained tasks, with limitations in addressing modality entanglement in fine-grained continual learning settings. To bridge this gap, we introduce a novel Continual Audio-Visual Segmentation (CAVS) task, aiming to continuously segment new classes guided by audio. Through comprehensive analysis, two critical challenges are identified: 1) multi-modal semantic drift, where a sounding objects is labeled as background in sequential tasks; 2) co-occurrence confusion, where frequent co-occurring classes tend to be confused. In this work, a Collision-based Multi-modal Rehearsal (CMR) framework is designed to address these challenges. Specifically, for multi-modal semantic drift, a Multi-modal Sample Selection (MSS) strategy is proposed to select samples with high modal consistency for rehearsal. Meanwhile, for co-occurence confusion, a Collision-based Sample Rehearsal (CSR) mechanism is designed, allowing for the increase of rehearsal sample frequency of those confusable classes during training process. Moreover, we construct three audio-visual incremental scenarios to verify effectiveness of our method. Comprehensive experiments demonstrate that our method significantly outperforms single-modal continual learning methods. Code can be seen at https://github.com/cqu-student/CAVS-CMR.