Publications
* Equal Contribution | † Corresponding Author
2025
- TCSVT
Yunlong Tang*, Jing Bi*, Siting Xu*, Luchuan Song, Susan Liang, Teng Wang, Daoan Zhang, Jie An, Jingyang Lin, Rongyi Zhu, and 10 more authorsIEEE Transactions on Circuits and Systems for Video Technology (TCSVT), 2025With the burgeoning growth of online video platforms and the escalating volume of video content, the demand for proficient video understanding tools has intensified markedly. Given the remarkable capabilities of large language models (LLMs) in language and multimodal tasks, this survey provides a detailed overview of recent advancements in video understanding that harness the power of LLMs (Vid-LLMs). The emergent capabilities of Vid-LLMs are surprisingly advanced, particularly their ability for open-ended multi-granularity (general, temporal, and spatiotemporal) reasoning combined with commonsense knowledge, suggesting a promising path for future video understanding. We examine the unique characteristics and capabilities of Vid-LLMs, categorizing the approaches into three main types: Video Analyzer x LLM, Video Embedder x LLM, and (Analyzer + Embedder) x LLM. Furthermore, we identify five sub-types based on the functions of LLMs in Vid-LLMs: LLM as Summarizer, LLM as Manager, LLM as Text Decoder, LLM as Regressor, and LLM as Hidden Layer. Furthermore, this survey presents a comprehensive study of the tasks, datasets, benchmarks, and evaluation methodologies for Vid-LLMs. Additionally, it explores the expansive applications of Vid-LLMs across various domains, highlighting their remarkable scalability and versatility in real-world video understanding challenges. Finally, it summarizes the limitations of existing Vid-LLMs and outlines directions for future research. For more information, readers are recommended to visit the repository at https://github.com/yunlong10/awesome-llms-for-video-understanding.
@article{vidllmsurvey, title = {Video Understanding with Large Language Models: A Survey 🔥🔥🔥}, author = {Tang, Yunlong and Bi, Jing and Xu, Siting and Song, Luchuan and Liang, Susan and Wang, Teng and Zhang, Daoan and An, Jie and Lin, Jingyang and Zhu, Rongyi and Vosoughi, Ali and Huang, Chao and Zhang, Zeliang and Liu, Pinxin and Feng, Mingqian and Zheng, Feng and Zhang, Jianguo and Luo, Ping and Luo, Jiebo and Xu, Chenliang}, journal = {IEEE Transactions on Circuits and Systems for Video Technology (TCSVT)}, year = {2025}, } - NeurIPS
Yunlong Tang, Pinxin Liu, Mingqian Feng, Zhangyun Tan, Rui Mao, Chao Huang, Jing Bi, Yunzhong Xiao, Susan Liang, Hang Hua, and 4 more authorsThe 39th Annual Conference on Neural Information Processing Systems (NeurIPS), 2025Understanding perspective is fundamental to human visual perception, yet the extent to which multimodal large language models (MLLMs) internalize perspective geometry remains unclear. We introduce MMPerspective, the first benchmark specifically designed to systematically evaluate MLLMs’ understanding of perspective through 10 carefully crafted tasks across three complementary dimensions: Perspective Perception, Reasoning, and Robustness. Our benchmark comprises 2,711 real-world and synthetic image instances with 5,083 question-answer pairs that probe key capabilities, such as vanishing point perception and counting, perspective type reasoning, line relationship understanding in 3D space, invariance to perspective-preserving transformations, etc. Through a comprehensive evaluation of 43 state-of-the-art MLLMs, we uncover significant limitations: while models demonstrate competence on surface-level perceptual tasks, they struggle with compositional reasoning and maintaining spatial consistency under perturbations. Our analysis further reveals intriguing patterns between model architecture, scale, and perspective capabilities, highlighting both robustness bottlenecks and the benefits of chain-of-thought prompting. MMPerspective establishes a valuable testbed for diagnosing and advancing spatial understanding in vision-language systems. Resources available at: https://yunlong10.github.io/MMPerspective/
@article{tang2025mmperspective, title = {MMPerspective: Do MLLMs Understand Perspective? A Comprehensive Benchmark for Perspective Perception, Reasoning, and Robustness}, author = {Tang, Yunlong and Liu, Pinxin and Feng, Mingqian and Tan, Zhangyun and Mao, Rui and Huang, Chao and Bi, Jing and Xiao, Yunzhong and Liang, Susan and Hua, Hang and Vosoughi, Ali and Song, Luchuan and Zhang, Zeliang and Xu, Chenliang}, journal = {The 39th Annual Conference on Neural Information Processing Systems (NeurIPS)}, year = {2025}, } - NeurIPS
Hang Hua, Ziyun Zeng, Yizhi Song, Yunlong Tang, Liu He, Daniel Aliaga, Wei Xiong, and Jiebo LuoThe 39th Annual Conference on Neural Information Processing Systems (NeurIPS), 2025Recent multimodal image generators such as GPT-4o, Gemini 2.0 Flash, and Gemini 2.5 Pro excel at following complex instructions, editing images, and maintaining concept consistency. However, they are still evaluated by disjoint toolkits: text-to-image (T2I) benchmarks that lack multi-modal conditioning, and customized image generation benchmarks that overlook compositional semantics and common knowledge. We propose MMIG-Bench, a comprehensive Multi-Modal Image Generation Benchmark that unifies these tasks by pairing 4,850 richly annotated text prompts with 1,750 multi-view reference images across 380 subjects, spanning humans, animals, objects, and artistic styles. MMIG-Bench is equipped with a three-level evaluation framework: (1) low-level metrics for visual artifacts and identity preservation of objects; (2) novel Aspect Matching Score (AMS): a VQA-based mid-level metric that delivers fine-grained prompt-image alignment and shows strong correlation with human judgments; and (3) high-level metrics for aesthetics and human preference. Using MMIG-Bench, we benchmark 17 state-of-the-art models, including Gemini 2.5 Pro, FLUX, DreamBooth, and IP-Adapter, and validate our metrics with 32k human ratings, yielding in-depth insights into architecture and data design.
@article{hua2025mmigbench, title = {MMIG-Bench: Towards Comprehensive and Explainable Evaluation of Multi-Modal Image Generation Models}, author = {Hua, Hang and Zeng, Ziyun and Song, Yizhi and Tang, Yunlong and He, Liu and Aliaga, Daniel and Xiong, Wei and Luo, Jiebo}, journal = {The 39th Annual Conference on Neural Information Processing Systems (NeurIPS)}, year = {2025}, } - NeurIPS
Chao Huang, Yuesheng Ma, Junxuan Huang, Susan Liang, Yunlong Tang, Jing Bi, Wenqiang Liu, Nima Mesgarani, and Chenliang XuThe 39th Annual Conference on Neural Information Processing Systems (NeurIPS), 2025Audio source separation is fundamental for machines to understand complex acoustic environments and underpins numerous audio applications. Current supervised deep learning approaches, while powerful, are limited by the need for extensive, task-specific labeled data and struggle to generalize to the immense variability and open-set nature of real-world acoustic scenes. Inspired by the success of generative foundation models, we investigate whether pre-trained text-guided audio diffusion models can overcome these limitations. We make a surprising discovery: zero-shot source separation can be achieved purely through a pre-trained text-guided audio diffusion model under the right configuration. Our method, named ZeroSep, works by inverting the mixed audio into the diffusion model’s latent space and then using text conditioning to guide the denoising process to recover individual sources. Without any task-specific training or fine-tuning, ZeroSep repurposes the generative diffusion model for a discriminative separation task and inherently supports open-set scenarios through its rich textual priors. ZeroSep is compatible with a variety of pre-trained text-guided audio diffusion backbones and delivers strong separation performance on multiple separation benchmarks, surpassing even supervised methods.
@article{huang2025zerosep, title = {ZeroSep: Separate Anything in Audio with Zero Training}, author = {Huang, Chao and Ma, Yuesheng and Huang, Junxuan and Liang, Susan and Tang, Yunlong and Bi, Jing and Liu, Wenqiang and Mesgarani, Nima and Xu, Chenliang}, journal = {The 39th Annual Conference on Neural Information Processing Systems (NeurIPS)}, year = {2025}, } - NeurIPS
Jiani Liu, Zhiyuan Wang, Zeliang Zhang, Chao Huang, Susan Liang, Yunlong Tang, and Chenliang XuThe 39th Annual Conference on Neural Information Processing Systems (NeurIPS), 2025@article{liu2025harnessing, title = {Harnessing the Computation Redundancy in ViTs to Boost Adversarial Transferability}, author = {Liu, Jiani and Wang, Zhiyuan and Zhang, Zeliang and Huang, Chao and Liang, Susan and Tang, Yunlong and Xu, Chenliang}, journal = {The 39th Annual Conference on Neural Information Processing Systems (NeurIPS)}, year = {2025}, } - CVPR
Yunlong Tang*, Junjia Guo*, Hang Hua, Susan Liang, Mingqian Feng, Xinyang Li, Rui Mao, Chao Huang, Jing Bi, Zeliang Zhang, and 2 more authorsIn Proceedings of the Computer Vision and Pattern Recognition Conference (CVPR), 2025The advancement of Multimodal Large Language Models (MLLMs) has enabled significant progress in multimodal understanding, expanding their capacity to analyze video content. However, existing evaluation benchmarks for MLLMs primarily focus on abstract video comprehension, lacking a detailed assessment of their ability to understand video compositions, the nuanced interpretation of how visual elements combine and interact within highly compiled video contexts. We introduce VidComposition, a new benchmark specifically designed to evaluate the video composition understanding capabilities of MLLMs using carefully curated compiled videos and cinematic-level annotations. VidComposition includes 982 videos with 1706 multiple-choice questions, covering various compositional aspects such as camera movement, angle, shot size, narrative structure, character actions and emotions, etc. Our comprehensive evaluation of 33 open-source and proprietary MLLMs reveals a significant performance gap between human and model capabilities. This highlights the limitations of current MLLMs in understanding complex, compiled video compositions and offers insights into areas for further improvement. The leaderboard and evaluation code are available at https://yunlong10.github.io/VidComposition/.
@inproceedings{tang2024vidcompostion, title = {VidComposition: Can MLLMs Analyze Compositions in Compiled Videos?}, author = {Tang, Yunlong and Guo, Junjia and Hua, Hang and Liang, Susan and Feng, Mingqian and Li, Xinyang and Mao, Rui and Huang, Chao and Bi, Jing and Zhang, Zeliang and Fazli, Pooyan and Xu, Chenliang}, booktitle = {Proceedings of the Computer Vision and Pattern Recognition Conference (CVPR)}, year = {2025}, pages = {8490-8500}, } - ICCVW
Yunlong Tang, Junjia Guo, Pinxin Liu, Zhiyuan Wang, Hang Hua, Jia-Xing Zhong, Yunzhong Xiao, Chao Huang, Luchuan Song, Susan Liang, and 7 more authorsarXiv preprint arXiv:2501.06250, 2025Traditional Celluloid (Cel) Animation production pipeline encompasses multiple essential steps, including storyboarding, layout design, keyframe animation, inbetweening, and colorization, which demand substantial manual effort, technical expertise, and significant time investment. These challenges have historically impeded the efficiency and scalability of Cel-Animation production. The rise of generative artificial intelligence (GenAI), encompassing large language models, multimodal models, and diffusion models, offers innovative solutions by automating tasks such as inbetween frame generation, colorization, and storyboard creation. This survey explores how GenAI integration is revolutionizing traditional animation workflows by lowering technical barriers, broadening accessibility for a wider range of creators through tools like AniDoc, ToonCrafter, and AniSora, and enabling artists to focus more on creative expression and artistic innovation. Despite its potential, issues such as maintaining visual consistency, ensuring stylistic coherence, and addressing ethical considerations continue to pose challenges. Furthermore, this paper discusses future directions and explores potential advancements in AI-assisted animation. For further exploration and resources, please visit our GitHub repository: https://github.com/yunlong10/Awesome-AI4Animation
@article{tang2025ai4anime, title = {Generative AI for Cel-Animation: A Survey 🔥🔥🔥}, author = {Tang, Yunlong and Guo, Junjia and Liu, Pinxin and Wang, Zhiyuan and Hua, Hang and Zhong, Jia-Xing and Xiao, Yunzhong and Huang, Chao and Song, Luchuan and Liang, Susan and Song, Yizhi and He, Liu and Bi, Jing and Feng, Mingqian and Li, Xinyang and Zhang, Zeliang and Xu, Chenliang}, journal = {arXiv preprint arXiv:2501.06250}, year = {2025}, } - CVPR
Jing Bi, Junjia Guo, Yunlong Tang, Lianggong Bruce Wen, Zhang Liu, and Chenliang XuIn Proceedings of the Computer Vision and Pattern Recognition Conference (CVPR), 2025Recent advancements in Multimodal Large Language Models (MLLMs) have demonstrated remarkable progress in visual understanding. This impressive leap raises a compelling question: how can language models, initially trained solely on linguistic data, effectively interpret and process visual content? This paper aims to address this question with systematic investigation across 4 model families and 4 model scales, uncovering a unique class of attention heads that focus specifically on visual content. Our analysis reveals a strong correlation between the behavior of these attention heads, the distribution of attention weights, and their concentration on visual tokens within the input. These findings enhance our understanding of how LLMs adapt to multimodal tasks, demonstrating their potential to bridge the gap between textual and visual understanding. This work paves the way for the development of AI systems capable of engaging with diverse modalities.
@inproceedings{bi2024unveiling, title = {Unveiling Visual Perception in Language Models: An Attention Head Analysis Approach}, author = {Bi, Jing and Guo, Junjia and Tang, Yunlong and Wen, Lianggong Bruce and Liu, Zhang and Xu, Chenliang}, booktitle = {Proceedings of the Computer Vision and Pattern Recognition Conference (CVPR)}, year = {2025}, } - AAAI
Yunlong Tang, Daiki Shimada, Jing Bi, Mingqian Feng, Hang Hua, and Chenliang Xu†In Proceedings of the AAAI Conference on Artificial Intelligence (AAAI), 2025Large language models (LLMs) have demonstrated remarkable capabilities in natural language and multimodal domains. By fine-tuning multimodal LLMs with temporal annotations from well-annotated datasets, e.g., dense video captioning datasets, their temporal understanding capacity in video-language tasks can be obtained. However, there is a notable lack of untrimmed audio-visual video datasets with precise temporal annotations for events. This deficiency hinders LLMs from learning the alignment between time, audio-visual events, and text tokens, thus impairing their ability to localize audio-visual events in videos temporally. To address this gap, we introduce PU-VALOR, a comprehensive audio-visual dataset comprising over 114,000 pseudo-untrimmed videos with detailed temporal annotations. PU-VALOR is derived from the large-scale but coarse-annotated audio-visual dataset VALOR, through a subtle method involving event-based video clustering, random temporal scaling, and permutation. By fine-tuning a multimodal LLM on PU-VALOR, we developed AVicuna, a model capable of aligning audio-visual events with temporal intervals and corresponding text tokens. AVicuna excels in temporal localization and time-aware dialogue capabilities. Our experiments demonstrate that AVicuna effectively handles temporal understanding in audio-visual videos and achieves state-of-the-art performance on open-ended video QA, audio-visual QA, and audio-visual event dense localization tasks.
@inproceedings{tang2024avicuna, title = {Empowering LLMs with Pseudo-Untrimmed Videos for Audio-Visual Temporal Understanding}, author = {Tang, Yunlong and Shimada, Daiki and Bi, Jing and Feng, Mingqian and Hua, Hang and Xu, Chenliang}, booktitle = {Proceedings of the AAAI Conference on Artificial Intelligence (AAAI)}, volume = {39}, number = {7}, pages = {7293-7301}, year = {2025}, doi = {10.1609/aaai.v39i7.32784}, } - AAAI
Yunlong Tang, Gen Zhan, Li Yang, Yiting Liao, and Chenliang Xu†In Proceedings of the AAAI Conference on Artificial Intelligence (AAAI), 2025Video saliency prediction aims to identify the regions in a video that attract human attention and gaze, driven by bottom-up features from the video and top-down processes like memory and cognition. Among these top-down influences, language plays a crucial role in guiding attention by shaping how visual information is interpreted. Existing methods primarily focus on modeling perceptual information while neglecting the reasoning process facilitated by language, where ranking cues are crucial outcomes of this process and practical guidance for saliency prediction. In this paper, we propose CaRDiff (Caption, Rank, and generate with Diffusion), a framework that imitates the process by integrating multimodal large language model (MLLM), a grounding module, and a diffusion model, to enhance video saliency prediction. Specifically, we introduce a novel prompting method VSOR-CoT (Video Salient Object Ranking Chain of Thought), which utilizes an MLLM with a grounding module to caption video content and infer salient objects along with their rankings and positions. This process derives ranking maps that can be sufficiently leveraged by the diffusion model to decode the saliency maps for the given video accurately. Extensive experiments show the effectiveness of VSOR-CoT in improving the performance of video saliency prediction. The proposed CaRDiff performs better than state-of-the-art models on the MVS dataset and demonstrates cross-dataset capabilities on the DHF1k dataset through zero-shot evaluation.
@inproceedings{tang2024cardiff, title = {CaRDiff: Video Salient Object Ranking Chain of Thought Reasoning for Saliency Prediction with Diffusion}, author = {Tang, Yunlong and Zhan, Gen and Yang, Li and Liao, Yiting and Xu, Chenliang}, booktitle = {Proceedings of the AAAI Conference on Artificial Intelligence (AAAI)}, year = {2025}, doi = {10.1609/aaai.v39i7.32785}, volume = {39}, number = {7}, pages = {7302-7310}, } - AAAI
Hang Hua*, Yunlong Tang*, Chenliang Xu, and Jiebo Luo†In Proceedings of the AAAI Conference on Artificial Intelligence (AAAI), 2025Video summarization aims to create short, accurate, and cohesive summaries of longer videos. Despite the existence of various video summarization datasets, a notable limitation is their limited amount of source videos, which hampers the effective training of advanced large vision-language models (VLMs). Additionally, most existing datasets are created for video-to-video summarization, overlooking the contemporary need for multimodal video content summarization. Recent efforts have been made to expand from unimodal to multimodal video summarization, categorizing the task into three sub-tasks based on the summary’s modality: video-to-video (V2V), video-to-text (V2T), and a combination of video and text summarization (V2VT). However, the textual summaries in previous multimodal datasets are inadequate. To address these issues, we introduce Instruct-V2Xum, a cross-modal video summarization dataset featuring 30,000 diverse videos sourced from YouTube, with lengths ranging from 40 to 940 seconds and an average summarization ratio of 16.39%. Each video summary in Instruct-V2Xum is paired with a textual summary that references specific frame indexes, facilitating the generation of aligned video and textual summaries. In addition, we propose a new video summarization framework named V2Xum-LLM. V2Xum-LLM, specifically V2Xum-LLaMA in this study, is the first framework that unifies different video summarization tasks into one large language model’s (LLM) text decoder and achieves task-controllable video summarization with temporal prompts and task instructions. Experiments show that V2Xum-LLaMA outperforms strong baseline models on multiple video summarization tasks. Furthermore, we propose an enhanced evaluation metric for V2V and V2VT summarization tasks.
@inproceedings{hua2024v2xum, title = {V2Xum-LLM: Cross-modal Video Summarization with Temporal Prompt Instruction Tuning}, author = {Hua, Hang and Tang, Yunlong and Xu, Chenliang and Luo, Jiebo}, booktitle = {Proceedings of the AAAI Conference on Artificial Intelligence (AAAI)}, volume = {39}, number = {4}, pages = {3599-3607}, year = {2025}, doi = {10.1609/aaai.v39i4.32374}, } - CVPRW
Chao Huang, Susan Liang, Yunlong Tang, Li Ma, Yapeng Tian, and Chenliang Xu†arXiv preprint arXiv:2504.02154, 2025Diffusion models offer impressive controllability for image tasks, primarily through noise predictions that encode task-specific information and classifier-free guidance enabling adjustable scaling. This scaling mechanism implicitly defines a “scaling space” whose potential for fine-grained semantic manipulation remains underexplored. We investigate this space, starting with inversion-based editing where the difference between conditional/unconditional noise predictions carries key semantic information. Our core contribution stems from a Fourier analysis of noise predictions, revealing that its low- and high-frequency components evolve differently throughout diffusion. Based on this insight, we introduce FreSca, a straightforward method that applies guidance scaling independently to different frequency bands in the Fourier domain. FreSca demonstrably enhances existing image editing methods without retraining. Excitingly, its effectiveness extends to image understanding tasks such as depth estimation, yielding quantitative gains across multiple datasets.
@article{huang2025fresca, title = {FreSca: Unveiling the Scaling Space in Diffusion Models}, author = {Huang, Chao and Liang, Susan and Tang, Yunlong and Ma, Li and Tian, Yapeng and Xu, Chenliang}, journal = {arXiv preprint arXiv:2504.02154}, year = {2025}, } - CVPRW
Liu He, Yizhi Song, Hejun Huang, Pinxin Liu, Yunlong Tang, Daniel Aliaga, and Xin ZhouarXiv preprint arXiv:2408.10453, 2025Text-to-video generation has been dominated by diffusion-based or autoregressive models. These novel models provide plausible versatility, but are criticized for improper physical motion, shading and illumination, camera motion, and temporal consistency. The film industry relies on manually-edited Computer-Generated Imagery (CGI) using 3D modeling software. Human-directed 3D synthetic videos address these shortcomings, but require tight collaboration between movie makers and 3D rendering experts. We introduce an automatic synthetic video generation pipeline based on Vision Large Language Model (VLM) agent collaborations. Given a language description of a video, multiple VLM agents direct various processes of the generation pipeline. They cooperate to create Blender scripts which render a video following the given description. Augmented with Blender-based movie making knowledge, the Director agent decomposes the text-based video description into sub-processes. For each sub-process, the Programmer agent produces Python-based Blender scripts based on function composing and API calling. The Reviewer agent, with knowledge of video reviewing, character motion coordinates, and intermediate screenshots, provides feedback to the Programmer agent. The Programmer agent iteratively improves scripts to yield the best video outcome. Our generated videos show better quality than commercial video generation models in five metrics on video quality and instruction-following performance. Our framework outperforms other approaches in a user study on quality, consistency, and rationality.
@article{kubrick2025, title = {Kubrick: Multimodal Agent Collaborations for Synthetic Video Generation}, author = {He, Liu and Song, Yizhi and Huang, Hejun and Liu, Pinxin and Tang, Yunlong and Aliaga, Daniel and Zhou, Xin}, journal = {arXiv preprint arXiv:2408.10453}, year = {2025}, } - 3DV
Pinxin Liu, Luchuan Song, Daoan Zhang, Hang Hua, Yunlong Tang, Huaijin Tu, Jiebo Luo, and Chenliang Xu†International Conference on 3D Vision (3DV), 2025Existing methods like Neural Radiation Fields (NeRF) and 3D Gaussian Splatting (3DGS) have made significant strides in facial attribute control such as facial animation and components editing, yet they struggle with fine-grained representation and scalability in dynamic head modeling. To address these limitations, we propose GaussianStyle, a novel framework that integrates the volumetric strengths of 3DGS with the powerful implicit representation of StyleGAN. The GaussianStyle preserves structural information, such as expressions and poses, using Gaussian points, while projecting the implicit volumetric representation into StyleGAN to capture high-frequency details and mitigate the over-smoothing commonly observed in neural texture rendering. Experimental outcomes indicate that our method achieves state-of-the-art performance in reenactment, novel view synthesis, and animation.
@article{liu2024emo, title = {GaussianStyle: Gaussian Head Avatar via StyleGAN}, author = {Liu, Pinxin and Song, Luchuan and Zhang, Daoan and Hua, Hang and Tang, Yunlong and Tu, Huaijin and Luo, Jiebo and Xu, Chenliang}, journal = {International Conference on 3D Vision (3DV)}, year = {2025}, } -
Yunlong Tang, Jing Bi, Chao Huang, Susan Liang, Daiki Shimada, Hang Hua, Yunzhong Xiao, Yizhi Song, Pinxin Liu, Mingqian Feng, and 9 more authorsarXiv preprint arXiv:2504.05541, 2025We present CAT-V (Caption AnyThing in Video), a training-free framework for fine-grained object-centric video captioning that enables detailed descriptions of user-selected objects through time. CAT-V integrates three key components: a Segmenter based on SAMURAI for precise object segmentation across frames, a Temporal Analyzer powered by TRACE-Uni for accurate event boundary detection and temporal analysis, and a Captioner using InternVL-2.5 for generating detailed object-centric descriptions. Through spatiotemporal visual prompts and chain-of-thought reasoning, our framework generates detailed, temporally-aware descriptions of objects’ attributes, actions, statuses, interactions, and environmental contexts without requiring additional training data. CAT-V supports flexible user interactions through various visual prompts (points, bounding boxes, and irregular regions) and maintains temporal sensitivity by tracking object states and interactions across different time segments. Our approach addresses limitations of existing video captioning methods, which either produce overly abstract descriptions or lack object-level precision, enabling fine-grained, object-specific descriptions while maintaining temporal coherence and spatial accuracy. The GitHub repository for this project is available at: https://github.com/yunlong10/CAT-V
@article{tang2025catv, title = {Caption Anything in Video: Fine-grained Object-centric Captioning via Spatiotemporal Multimodal Prompting}, author = {Tang, Yunlong and Bi, Jing and Huang, Chao and Liang, Susan and Shimada, Daiki and Hua, Hang and Xiao, Yunzhong and Song, Yizhi and Liu, Pinxin and Feng, Mingqian and Guo, Junjia and Liu, Zhuo and Song, Luchuan and Vosoughi, Ali and He, Jinxi and He, Liu and Zhang, Zeliang and Luo, Jiebo and Xu, Chenliang}, journal = {arXiv preprint arXiv:2504.05541}, year = {2025}, } -
Jing Bi, Susan Liang, Xiaofei Zhou, Pinxin Liu, Junjia Guo, Yunlong Tang, Luchuan Song, Chao Huang, Guangyu Sun, Jinxi He, and 8 more authorsarXiv preprint arXiv:2504.03151, 2025Reasoning is central to human intelligence, enabling structured problem-solving across diverse tasks. Recent advances in large language models (LLMs) have greatly enhanced their reasoning abilities in arithmetic, commonsense, and symbolic domains. However, effectively extending these capabilities into multimodal contexts-where models must integrate both visual and textual inputs-continues to be a significant challenge. Multimodal reasoning introduces complexities, such as handling conflicting information across modalities, which require models to adopt advanced interpretative strategies. Addressing these challenges involves not only sophisticated algorithms but also robust methodologies for evaluating reasoning accuracy and coherence. This paper offers a concise yet insightful overview of reasoning techniques in both textual and multimodal LLMs. Through a thorough and up-to-date comparison, we clearly formulate core reasoning challenges and opportunities, highlighting practical methods for post-training optimization and test-time inference. Our work provides valuable insights and guidance, bridging theoretical frameworks and practical implementations, and sets clear directions for future research.
-
Jing Bi, Junjia Guo, Susan Liang, Guangyu Sun, Luchuan Song, Yunlong Tang, Jinxi He, Jiarui Wu, Ali Vosoughi, Chen Chen, and 1 more authorarXiv preprint arXiv:2503.11557, 2025Visual reasoning is central to human cognition, enabling individuals to interpret and abstractly understand their environment. Although recent Multimodal Large Language Models (MLLMs) have demonstrated impressive performance across language and vision-language tasks, existing benchmarks primarily measure recognition-based skills and inadequately assess true visual reasoning capabilities. To bridge this critical gap, we introduce VERIFY, a benchmark explicitly designed to isolate and rigorously evaluate the visual reasoning capabilities of state-of-the-art MLLMs. VERIFY compels models to reason primarily from visual information, providing minimal textual context to reduce reliance on domain-specific knowledge and linguistic biases. Each problem is accompanied by a human-annotated reasoning path, making it the first to provide in-depth evaluation of model decision-making processes. Additionally, we propose novel metrics that assess visual reasoning fidelity beyond mere accuracy, highlighting critical imbalances in current model reasoning patterns. Our comprehensive benchmarking of leading MLLMs uncovers significant limitations, underscoring the need for a balanced and holistic approach to both perception and reasoning. For more teaser and testing, visit our project page at https://verify-eqh.pages.dev/
@article{bi2025verify, title = {VERIFY: A Benchmark of Visual Explanation and Reasoning for Investigating Multimodal Reasoning Fidelity}, author = {Bi, Jing and Guo, Junjia and Liang, Susan and Sun, Guangyu and Song, Luchuan and Tang, Yunlong and He, Jinxi and Wu, Jiarui and Vosoughi, Ali and Chen, Chen and Xu, Chenliang}, journal = {arXiv preprint arXiv:2503.11557}, year = {2025}, }
2024
- ACM MM
Jing Bi, Yunlong Tang, Luchuan Song, Ali Vosoughi, Nguyen Nguyen, and Chenliang Xu†In Proceedings of the 32nd ACM International Conference on Multimedia (ACM MM), 2024The rapid evolution of egocentric video analysis brings new insights into understanding human activities and intentions from a first-person perspective. Despite this progress, the fragmentation in tasks like action recognition, procedure learning, and moment retrieval, etc., coupled with inconsistent annotations and isolated model development, hinders a holistic interpretation of video content. In response, we introduce the EAGLE (Egocentric AGgregated Language-video Engine) model and the EAGLE-400K dataset to provide a unified framework that integrates various egocentric video understanding tasks. EAGLE-400K, the first large-scale instruction-tuning dataset tailored for egocentric video, features 400K diverse samples to enhance a broad spectrum task from activity recognition to procedure knowledge learning. Moreover, EAGLE, a strong video-based multimodal large language model (MLLM), is designed to effectively capture both spatial and temporal information. In addition, we propose a set of evaluation metrics designed to facilitate a thorough assessment of MLLM for egocentric video understanding. Our extensive experiments demonstrate EAGLE’s superior performance over existing models, highlighting its ability to balance task-specific understanding with comprehensive video interpretation. With EAGLE, we aim to pave the way for novel research opportunities and practical applications in real-world scenarios.
@inproceedings{bi2024eagle, title = {EAGLE: Egocentric AGgregated Language-video Engine}, author = {Bi, Jing and Tang, Yunlong and Song, Luchuan and Vosoughi, Ali and Nguyen, Nguyen and Xu, Chenliang}, booktitle = {Proceedings of the 32nd ACM International Conference on Multimedia (ACM MM)}, year = {2024}, } - ECCVW
Andrey Moskalenko, Alexey Bryncev, Dmitry Vatolin, Radu Timofte, Gen Zhan, Li Yang, Yunlong Tang, Yiting Liao, Jiongzhi Lin, Baitao Huang, and 23 more authorsIn Proceedings of the European Conference on Computer Vision (ECCV) Workshops, 2024This paper reviews the Challenge on Video Saliency Prediction at AIM 2024. The goal of the participants was to develop a method for predicting accurate saliency maps for the provided set of video sequences. Saliency maps are widely exploited in various applications, including video compression, quality assessment, visual perception studies, the advertising industry, etc. For this competition, a previously unused large-scale audio-visual mouse saliency (AViMoS) dataset of 1500 videos with more than 70 observers per video was collected using crowdsourced mouse tracking. The dataset collection methodology has been validated using conventional eye-tracking data and has shown high consistency. Over 30 teams registered in the challenge, and there are 7 teams that submitted the results in the final phase. The final phase solutions were tested and ranked by commonly used quality metrics on a private test subset. The results of this evaluation and the descriptions of the solutions are presented in this report. All data, including the private test subset, is made publicly available on the challenge homepage - https://challenges.videoprocessing.ai/challenges/video-saliency-prediction.html
@inproceedings{moskalenko2024aim, title = {AIM 2024 Challenge on Video Saliency Prediction: Methods and Results}, author = {Moskalenko, Andrey and Bryncev, Alexey and Vatolin, Dmitry and Timofte, Radu and Zhan, Gen and Yang, Li and Tang, Yunlong and Liao, Yiting and Lin, Jiongzhi and Huang, Baitao and Moradi, Morteza and Moradi, Mohammad and Rundo, Francesco and Spampinato, Concetto and Borji, Ali and Palazzo, Simone and Zhu, Yuxin and Sun, Yinan and Duan, Huiyu and Cao, Yuqin and Jia, Ziheng and Hu, Qiang and Min, Xiongkuo and Zhai, Guangtao and Fang, Hao and Cong, Runmin and Lu, Xiankai and Zhou, Xiaofei and Zhang, Wei and Zhao, Chunyu and Mu, Wentao and Deng, Tao and Tavakoli, Hamed R}, booktitle = {Proceedings of the European Conference on Computer Vision (ECCV) Workshops}, year = {2024}, } -
Hang Hua*, Yunlong Tang*, Ziyun Zeng*, Liangliang Cao, Zhengyuan Yang, Hangfeng He, Chenliang Xu, and Jiebo LuoarXiv preprint arXiv:2410.09733, 2024The advent of large Vision-Language Models (VLMs) has significantly advanced multimodal understanding, enabling more sophisticated and accurate integration of visual and textual information across various tasks, including image and video captioning, visual question answering, and cross-modal retrieval. Despite VLMs’ superior capabilities, researchers lack a comprehensive understanding of their compositionality – the ability to understand and produce novel combinations of known visual and textual components. Prior benchmarks provide only a relatively rough compositionality evaluation from the perspectives of objects, relations, and attributes while neglecting deeper reasoning about object interactions, counting, and complex compositions. However, compositionality is a critical ability that facilitates coherent reasoning and understanding across modalities for VLMs. To address this limitation, we propose MMCOMPOSITION, a novel human-annotated benchmark for comprehensively and accurately evaluating VLMs’ compositionality. Our proposed benchmark serves as a complement to these earlier works. With MMCOMPOSITION, we can quantify and explore the compositionality of the mainstream VLMs. Surprisingly, we find GPT-4o’s compositionality inferior to the best open-source model, and we analyze the underlying reasons. Our experimental analysis reveals the limitations of VLMs in fine-grained compositional perception and reasoning, and points to areas for improvement in VLM design and training. Resources available at: https://hanghuacs.github.io/MMComposition/
@article{hua2024mmcomposition, title = {MMCOMPOSITION: Revisiting the Compositionality of Pre-trained Vision-Language Models}, author = {Hua, Hang and Tang, Yunlong and Zeng, Ziyun and Cao, Liangliang and Yang, Zhengyuan and He, Hangfeng and Xu, Chenliang and Luo, Jiebo}, journal = {arXiv preprint arXiv:2410.09733}, year = {2024}, } -
Mingqian Feng, Yunlong Tang, Zeliang Zhang, and Chenliang Xu†arXiv preprint arXiv:2406.12663, 2024Large Vision-Language Models (LVLMs) excel in integrating visual and linguistic contexts to produce detailed content, facilitating applications such as image captioning. However, using LVLMs to generate descriptions often faces the challenge of object hallucination (OH), where the output text misrepresents actual objects in the input image. While previous studies attribute the occurrence of OH to the inclusion of more details, our study finds technical flaws in existing metrics, leading to unreliable evaluations of models and conclusions about OH. This has sparked a debate on the question: Do more details always introduce more hallucinations in LVLM-based image captioning? In this paper, we address this debate by proposing a novel decoding strategy, Differentiated Beam Decoding (DBD), along with a reliable new set of evaluation metrics: CLIP-Precision, CLIP-Recall, and CLIP-F1. DBD decodes the wealth of information hidden in visual input into distinct language representations called unit facts in parallel. This decoding is achieved via a well-designed differential score that guides the parallel search and candidate screening. The selected unit facts are then aggregated to generate the final caption. Our proposed metrics evaluate the comprehensiveness and accuracy of image captions by comparing the embedding groups of ground-truth image regions and generated text partitions. Extensive experiments on the Visual Genome dataset validate the effectiveness of our approach, demonstrating that it produces detailed descriptions while maintaining low hallucination levels.
@article{feng2024more, title = {Do More Details Always Introduce More Hallucinations in LVLM-based Image Captioning?}, author = {Feng, Mingqian and Tang, Yunlong and Zhang, Zeliang and Xu, Chenliang}, journal = {arXiv preprint arXiv:2406.12663}, year = {2024}, } -
Chao Huang, Susan Liang, Yunlong Tang, Yapeng Tian, Anurag Kumar, and Chenliang Xu†arXiv preprint arXiv:2410.24151, 2024Text-guided diffusion models have revolutionized generative tasks by producing high-fidelity content from text descriptions. They have also enabled an editing paradigm where concepts can be replaced through text conditioning (e.g., a dog to a tiger). In this work, we explore a novel approach: instead of replacing a concept, can we enhance or suppress the concept itself? Through an empirical study, we identify a trend where concepts can be decomposed in text-guided diffusion models. Leveraging this insight, we introduce ScalingConcept, a simple yet effective method to scale decomposed concepts up or down in real input without introducing new elements. To systematically evaluate our approach, we present the WeakConcept-10 dataset, where concepts are imperfect and need to be enhanced. More importantly, ScalingConcept enables a variety of novel zero-shot applications across image and audio domains, including tasks such as canonical pose generation and generative sound highlighting or removal.
@article{huang2024scalingconcept, title = {Scaling Concept with Text-Guided Diffusion Models}, author = {Huang, Chao and Liang, Susan and Tang, Yunlong and Tian, Yapeng and Kumar, Anurag and Xu, Chenliang}, journal = {arXiv preprint arXiv:2410.24151}, year = {2024}, }
2023
- CVPRW
Yunlong Tang, Jinrui Zhang, Xiangchen Wang, Teng Wang, and Feng Zheng†IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW), 2023Our winning entry for the CVPR 2023 Generic Event Boundary Captioning (GEBC) competition is detailed in this paper. Unlike conventional video captioning tasks, GEBC demands that the captioning model possess an understanding of immediate changes in status around the designated video boundary, making it a difficult task. This paper proposes an effective model LLMVA-GEBC (Large Language Model with Video Adapter for Generic Event Boundary Captioning): (1) We utilize a pretrained LLM for generating human-like captions with high quality. (2) To adapt the model to the GEBC task, we take the video Q-former as an adapter and train it with the frozen visual feature extractors and LLM. Our proposed method achieved a 76.14 score on the test set and won the first place in the challenge. Our code is available at https://github.com/zjr2000/LLMVA-GEBC.
@article{tang2023llmva, title = {LLMVA-GEBC: Large Language Model with Video Adapter for Generic Event Boundary Captioning}, author = {Tang, Yunlong and Zhang, Jinrui and Wang, Xiangchen and Wang, Teng and Zheng, Feng}, journal = {IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW)}, year = {2023}, } - ICMC
Siting Xu*, Yunlong Tang*, and Feng Zheng†In Proceedings of the International Computer Music Conference (ICMC), 2023Launchpad is a popular music instrument for live performance and music production. In this paper, we propose LaunchpadGPT, a language model that can generate music visualization designs for Launchpad. We train the model on a large-scale dataset of music visualization designs and demonstrate its effectiveness in generating diverse and creative designs. We also develop a web-based platform that allows users to interact with the model and generate music visualization designs in real-time. Our code is available at https://github.com/yunlong10/LaunchpadGPT.
@inproceedings{xu2023launchpadgpt, title = {LaunchpadGPT: Language Model as Music Visualization Designer on Launchpad}, author = {Xu, Siting and Tang, Yunlong and Zheng, Feng}, booktitle = {Proceedings of the International Computer Music Conference (ICMC)}, year = {2023}, pages = {213-217}, } -
Teng Wang*, Jinrui Zhang*, Junjie Fei*, Hao Zheng, Yunlong Tang, Zhe Li, Mingqi Gao, and Shanshan ZhaoarXiv preprint arXiv:2305.02677, 2023Controllable image captioning is an emerging multimodal topic that aims to describe the image with natural language following human purpose, e.g., looking at the specified regions or telling in a particular text style. State-of-the-art methods are trained on annotated pairs of input controls and output captions. However, the scarcity of such well-annotated multimodal data largely limits their usability and scalability for interactive AI systems. Leveraging unimodal instruction-following foundation models is a promising alternative that benefits from broader sources of data. In this paper, we present Caption AnyThing (CAT), a foundation model augmented image captioning framework supporting a wide range of multimodel controls: 1) visual controls, including points, boxes, and trajectories; 2) language controls, such as sentiment, length, language, and factuality. Powered by Segment Anything Model (SAM) and ChatGPT, we unify the visual and language prompts into a modularized framework, enabling the flexible combination between different controls. Extensive case studies demonstrate the user intention alignment capabilities of our framework, shedding light on effective user interaction modeling in vision-language applications. Our code is publicly available at https://github.com/ttengwang/caption-anything
@article{wang2023caption, title = {Caption Anything: Interactive Image Description with Diverse Multimodal Controls}, author = {Wang, Teng and Zhang, Jinrui and Fei, Junjie and Zheng, Hao and Tang, Yunlong and Li, Zhe and Gao, Mingqi and Zhao, Shanshan}, journal = {arXiv preprint arXiv:2305.02677}, year = {2023}, }
2022
- ACCV
Yunlong Tang, Siting Xu, Teng Wang, Qin Lin, Qinglin Lu, and Feng Zheng†In Proceedings of the Asian Conference on Computer Vision (ACCV), 2022Advertisement video editing aims to automatically edit advertising videos into shorter videos while retaining coherent content and crucial information conveyed by advertisers. It mainly contains two stages: video segmentation and segment assemblage. The existing method performs well at video segmentation stages but suffers from the problems of dependencies on extra cumbersome models and poor performance at the segment assemblage stage. To address these problems, we propose M-SAN (Multi-modal Segment Assemblage Network) which can perform efficient and coherent segment assemblage task end-to-end. It utilizes multi-modal representation extracted from the segments and follows the Encoder-Decoder Ptr-Net framework with the Attention mechanism. Importance-coherence reward is designed for training M-SAN. We experiment on the Ads-1k dataset with 1000+ videos under rich ad scenarios collected from advertisers. To evaluate the methods, we propose a unified metric, Imp-Coh@Time, which comprehensively assesses the importance, coherence, and duration of the outputs at the same time. Experimental results show that our method achieves better performance than random selection and the previous method on the metric. Ablation experiments further verify that multi-modal representation and importance-coherence reward significantly improve the performance. Ads-1k dataset is available at: https://github.com/yunlong10/Ads-1k.
@inproceedings{Tang_2022_ACCV, author = {Tang, Yunlong and Xu, Siting and Wang, Teng and Lin, Qin and Lu, Qinglin and Zheng, Feng}, title = {Multi-modal Segment Assemblage Network for Ad Video Editing with Importance-Coherence Reward}, booktitle = {Proceedings of the Asian Conference on Computer Vision (ACCV)}, year = {2022}, pages = {3519-3535}, }
