Video understanding, including video captioning and retrieval, is still a great challenge for video-language models (VLMs). The existing video retrieval and caption benchmarks only include short descriptions, limits their ability of detailed video understanding evaluation. To address this problem, we present CaReBench, a testing benchmark for fine-grained video captioning and retrieval with 1,000 high-quality pairs of videos and human-annotated detailed captions. Uniquely, it provides manually separated spatial annotations and temporal annotations for each video. Based on this design, we introduce two evaluation metrics, ReBias and CapST, specifically tailored for video retrieval and video captioning tasks, respectively. These metrics enable a comprehensive investigation into the spatial and temporal biases inherent in VLMs. In addition, to handle both video retrieval and video captioning tasks in a unified framework, we develop a simple baseline based on a Multimodal Language Model (MLLM). By implementing a two-stage Supervised Fine-Tuning (SFT), we fully unlock the potential of MLLM, enabling it not only to generate detailed video descriptions but also to extract video features. Surprisingly, experimental results demonstrate that, compared to the CLIP-based models designed for retrieval and the popular MLLMs skilled in video captioning, our baseline shows competitive performance in both fine-grained video retrieval and video detailed captioning.
Understanding human-to-human interactions, especially in contexts like public security surveillance, is critical for monitoring and maintaining safety. Traditional activity recognition systems are limited by fixed vocabularies, predefined labels, and rigid interaction categories that often rely on choreographed videos and overlook concurrent interactive groups. These limitations make such systems less adaptable to real-world scenarios, where interactions are diverse and unpredictable. In this paper, we propose an open vocabulary human-to-human interaction recognition (OV-HHIR) framework that leverages large language models to generate open-ended textual descriptions of both seen and unseen human interactions in open-world settings without being confined to a fixed vocabulary. Additionally, we create a comprehensive, large-scale human-to-human interaction dataset by standardizing and combining existing public human interaction datasets into a unified benchmark. Extensive experiments demonstrate that our method outperforms traditional fixed-vocabulary classification systems and existing cross-modal language models for video understanding, setting the stage for more intelligent and adaptable visual understanding systems in surveillance and beyond.
One of the challenging tasks in the field of video understanding is extracting semantic content from video inputs. Most existing systems use language models to describe videos in natural language sentences, but this has several major shortcomings. Such systems can rely too heavily on the language model component and base their output on statistical regularities in natural language text rather than on the visual contents of the video. Additionally, natural language annotations cannot be readily processed by a computer, are difficult to evaluate with performance metrics and cannot be easily translated into a different natural language. In this paper, we propose a method to annotate videos with knowledge graphs, and so avoid these problems. Specifically, we propose a deep-learning-based model for this task that first predicts pairs of individuals and then the relations between them. Additionally, we propose an extension of our model for the inclusion of background knowledge in the construction of knowledge graphs.
Video Large Language Models (VideoLLMs) have made significant strides in video understanding but struggle with long videos due to the limitations of their backbone LLMs. Existing solutions rely on length extrapolation, which is memory-constrained, or visual token compression, which primarily leverages low-level temporal redundancy while overlooking the more effective high-level knowledge redundancy. To address this, we propose $\textbf{ReTaKe}$, a training-free method with two novel modules DPSelect and PivotKV, to jointly reduce both temporal visual redundancy and knowledge redundancy for video compression. To align with the way of human temporal perception, DPSelect identifies keyframes based on inter-frame distance peaks. To leverage LLMs' learned prior knowledge, PivotKV marks the keyframes as pivots and compress non-pivot frames by pruning low-attention tokens in their KV cache. ReTaKe enables VideoLLMs to process 8 times longer frames (up to 2048), outperforming similar-sized models by 3-5% and even rivaling much larger ones on VideoMME, MLVU, LongVideoBench, and LVBench. Moreover, by overlapping compression operations with prefilling, ReTaKe introduces only ~10% prefilling latency overhead while reducing decoding latency by ~20%. Our code is available at https://github.com/SCZwangxiao/video-ReTaKe.
Video Question Answering (Video QA) is a challenging video understanding task that requires models to comprehend entire videos, identify the most relevant information based on contextual cues from a given question, and reason accurately to provide answers. Recent advancements in Multimodal Large Language Models (MLLMs) have transformed video QA by leveraging their exceptional commonsense reasoning capabilities. This progress is largely driven by the effective alignment between visual data and the language space of MLLMs. However, for video QA, an additional space-time alignment poses a considerable challenge for extracting question-relevant information across frames. In this work, we investigate diverse temporal modeling techniques to integrate with MLLMs, aiming to achieve question-guided temporal modeling that leverages pre-trained visual and textual alignment in MLLMs. We propose T-Former, a novel temporal modeling method that creates a question-guided temporal bridge between frame-wise visual perception and the reasoning capabilities of LLMs. Our evaluation across multiple video QA benchmarks demonstrates that T-Former competes favorably with existing temporal modeling approaches and aligns with recent advancements in video QA.
Multimodal Large Language Models (MLLMs) have driven major advances in video understanding, yet their vulnerability to adversarial tampering and manipulations remains underexplored. To address this gap, we introduce MVTamperBench, a benchmark that systematically evaluates MLLM robustness against five prevalent tampering techniques: rotation, masking, substitution, repetition, and dropping. Built from 3.4K original videos-expanded to over 17K tampered clips spanning 19 video tasks. MVTamperBench challenges models to detect manipulations in spatial and temporal coherence. We evaluate 45 recent MLLMs from 15+ model families, revealing substantial variability in resilience across tampering types and showing that larger parameter counts do not necessarily guarantee robustness. MVTamperBench sets a new benchmark for developing tamper-resilient MLLM in safety-critical applications, including detecting clickbait, preventing harmful content distribution, and enforcing policies on media platforms. We release all code and data to foster open research in trustworthy video understanding. Code: https://amitbcp.github.io/MVTamperBench/ Data: https://huggingface.co/datasets/Srikant86/MVTamperBench
Video question answering (VideoQA) aims to answer natural language questions according to the given videos. Although existing models perform well in the factoid VideoQA task, they still face challenges in deep video understanding (DVU) task, which focuses on story videos. Compared to factoid videos, the most significant feature of story videos is storylines, which are composed of complex interactions and long-range evolvement of core story topics including characters, actions and locations. Understanding these topics requires models to possess DVU capability. However, existing DVU datasets rarely organize questions according to these story topics, making them difficult to comprehensively assess VideoQA models' DVU capability of complex storylines. Additionally, the question quantity and video length of these dataset are limited by high labor costs of handcrafted dataset building method. In this paper, we devise a large language model based multi-agent collaboration framework, StoryMind, to automatically generate a new large-scale DVU dataset. The dataset, FriendsQA, derived from the renowned sitcom Friends with an average episode length of 1,358 seconds, contains 44.6K questions evenly distributed across 14 fine-grained topics. Finally, We conduct comprehensive experiments on 10 state-of-the-art VideoQA models using the FriendsQA dataset.
It is significantly challenging to recognize daily human actions in homes due to the diversity and dynamic changes in unconstrained home environments. It spurs the need to continually adapt to various users and scenes. Fine-tuning current video understanding models on newly encountered domains often leads to catastrophic forgetting, where the models lose their ability to perform well on previously learned scenarios. To address this issue, we formalize the problem of Video Domain Incremental Learning (VDIL), which enables models to learn continually from different domains while maintaining a fixed set of action classes. Existing continual learning research primarily focuses on class-incremental learning, while the domain incremental learning has been largely overlooked in video understanding. In this work, we introduce a novel benchmark of domain incremental human action recognition for unconstrained home environments. We design three domain split types (user, scene, hybrid) to systematically assess the challenges posed by domain shifts in real-world home settings. Furthermore, we propose a baseline learning strategy based on replay and reservoir sampling techniques without domain labels to handle scenarios with limited memory and task agnosticism. Extensive experimental results demonstrate that our simple sampling and replay strategy outperforms most existing continual learning methods across the three proposed benchmarks.
In the domain of Multimodal Large Language Models (MLLMs), achieving human-centric video understanding remains a formidable challenge. Existing benchmarks primarily emphasize object and action recognition, often neglecting the intricate nuances of human emotions, behaviors, and speech visual alignment within video content. We present HumanVBench, an innovative benchmark meticulously crafted to bridge these gaps in the evaluation of video MLLMs. HumanVBench comprises 17 carefully designed tasks that explore two primary dimensions: inner emotion and outer manifestations, spanning static and dynamic, basic and complex, as well as single-modal and cross-modal aspects. With two advanced automated pipelines for video annotation and distractor-included QA generation, HumanVBench utilizes diverse state-of-the-art (SOTA) techniques to streamline benchmark data synthesis and quality assessment, minimizing human annotation dependency tailored to human-centric multimodal attributes. A comprehensive evaluation across 16 SOTA video MLLMs reveals notable limitations in current performance, especially in cross-modal and temporal alignment, underscoring the necessity for further refinement toward achieving more human-like understanding. HumanVBench is open-sourced to facilitate future advancements and real-world applications in video MLLMs.
In this paper, we introduce PruneVid, a visual token pruning method designed to enhance the efficiency of multi-modal video understanding. Large Language Models (LLMs) have shown promising performance in video tasks due to their extended capabilities in comprehending visual modalities. However, the substantial redundancy in video data presents significant computational challenges for LLMs. To address this issue, we introduce a training-free method that 1) minimizes video redundancy by merging spatial-temporal tokens, and 2) leverages LLMs' reasoning capabilities to selectively prune visual features relevant to question tokens, enhancing model efficiency. We validate our method across multiple video benchmarks, which demonstrate that PruneVid can prune over 80% of tokens while maintaining competitive performance combined with different model networks. This highlights its superior effectiveness and efficiency compared to existing pruning methods. Code: https://github.com/Visual-AI/PruneVid.
Boosted by Multi-modal Large Language Models (MLLMs), text-guided universal segmentation models for the image and video domains have made rapid progress recently. However, these methods are often developed separately for specific domains, overlooking the similarities in task settings and solutions across these two areas. In this paper, we define the union of referring segmentation and reasoning segmentation at both the image and video levels as Instructed Visual Segmentation (IVS). Correspondingly, we propose InstructSeg, an end-to-end segmentation pipeline equipped with MLLMs for IVS. Specifically, we employ an object-aware video perceiver to extract temporal and object information from reference frames, facilitating comprehensive video understanding. Additionally, we introduce vision-guided multi-granularity text fusion to better integrate global and detailed text information with fine-grained visual guidance. By leveraging multi-task and end-to-end training, InstructSeg demonstrates superior performance across diverse image and video segmentation tasks, surpassing both segmentation specialists and MLLM-based methods with a single model. Our code is available at https://github.com/congvvc/InstructSeg.
Text-guided Video Temporal Grounding (VTG) aims to localize relevant segments in untrimmed videos based on textual descriptions, encompassing two subtasks: Moment Retrieval (MR) and Highlight Detection (HD). Although previous typical methods have achieved commendable results, it is still challenging to retrieve short video moments. This is primarily due to the reliance on sparse and limited decoder queries, which significantly constrain the accuracy of predictions. Furthermore, suboptimal outcomes often arise because previous methods rank predictions based on isolated predictions, neglecting the broader video context. To tackle these issues, we introduce FlashVTG, a framework featuring a Temporal Feature Layering (TFL) module and an Adaptive Score Refinement (ASR) module. The TFL module replaces the traditional decoder structure to capture nuanced video content variations across multiple temporal scales, while the ASR module improves prediction ranking by integrating context from adjacent moments and multi-temporal-scale features. Extensive experiments demonstrate that FlashVTG achieves state-of-the-art performance on four widely adopted datasets in both MR and HD. Specifically, on the QVHighlights dataset, it boosts mAP by 5.8% for MR and 3.3% for HD. For short-moment retrieval, FlashVTG increases mAP to 125% of previous SOTA performance. All these improvements are made without adding training burdens, underscoring its effectiveness. Our code is available at https://github.com/Zhuo-Cao/FlashVTG.
Large language models (LLMs) have revolutionized video-based computer vision applications, including action recognition, anomaly detection, and video summarization. Videos inherently pose unique challenges, combining spatial complexity with temporal dynamics that are absent in static images or textual data. Current approaches to video understanding with LLMs often rely on pretrained video encoders to extract spatiotemporal features and text encoders to capture semantic meaning. These representations are integrated within LLM frameworks, enabling multimodal reasoning across diverse video tasks. However, the critical question persists: Can LLMs truly understand the concept of time, and how effectively can they reason about temporal relationships in videos? This work critically examines the role of LLMs in video processing, with a specific focus on their temporal reasoning capabilities. We identify key limitations in the interaction between LLMs and pretrained encoders, revealing gaps in their ability to model long-term dependencies and abstract temporal concepts such as causality and event progression. Furthermore, we analyze challenges posed by existing video datasets, including biases, lack of temporal annotations, and domain-specific limitations that constrain the temporal understanding of LLMs. To address these gaps, we explore promising future directions, including the co-evolution of LLMs and encoders, the development of enriched datasets with explicit temporal labels, and innovative architectures for integrating spatial, temporal, and semantic reasoning. By addressing these challenges, we aim to advance the temporal comprehension of LLMs, unlocking their full potential in video analysis and beyond.
Recently, multi-modal large language models have made significant progress. However, visual information lacking of guidance from the user's intention may lead to redundant computation and involve unnecessary visual noise, especially in long, untrimmed videos. To address this issue, we propose FocusChat, a text-guided multi-modal large language model (LLM) that emphasizes visual information correlated to the user's prompt. In detail, Our model first undergoes the semantic extraction module, which comprises a visual semantic branch and a text semantic branch to extract image and text semantics, respectively. The two branches are combined using the Spatial-Temporal Filtering Module (STFM). STFM enables explicit spatial-level information filtering and implicit temporal-level feature filtering, ensuring that the visual tokens are closely aligned with the user's query. It lowers the essential number of visual tokens inputted into the LLM. FocusChat significantly outperforms Video-LLaMA in zero-shot experiments, using an order of magnitude less training data with only 16 visual tokens occupied. It achieves results comparable to the state-of-the-art in few-shot experiments, with only 0.72M pre-training data.
Existing works on human-centric video understanding typically focus on analyzing specific moment or entire videos. However, many applications require higher precision at the frame level. In this work, we propose a novel task, BestShot, which aims to locate highlight frames within human-centric videos via language queries. This task demands not only a deep semantic comprehension of human actions but also precise temporal localization. To support this task, we introduce the BestShot Benchmark. %The benchmark is meticulously constructed by combining human detection and tracking, potential frame selection based on human judgment, and detailed textual descriptions crafted by human input to ensure precision. The benchmark is meticulously constructed by combining human-annotated highlight frames, detailed textual descriptions and duration labeling. These descriptions encompass three critical elements: (1) Visual content; (2) Fine-grained action; and (3) Human Pose Description. Together, these elements provide the necessary precision to identify the exact highlight frames in videos. To tackle this problem, we have collected two distinct datasets: (i) ShotGPT4o Dataset, which is algorithmically generated by GPT-4o and (ii) Image-SMPLText Dataset, a dataset with large-scale and accurate per-frame pose description leveraging PoseScript and existing pose estimation datasets. Based on these datasets, we present a strong baseline model, ShotVL, fine-tuned from InternVL, specifically for BestShot. We highlight the impressive zero-shot capabilities of our model and offer comparative analyses with existing SOTA models. ShotVL demonstrates a significant 52% improvement over InternVL on the BestShot Benchmark and a notable 57% improvement on the THUMOS14 Benchmark, all while maintaining the SOTA performance in general image classification and retrieval.
Long video understanding poses unique challenges due to their temporal complexity and low information density. Recent works address this task by sampling numerous frames or incorporating auxiliary tools using LLMs, both of which result in high computational costs. In this work, we introduce a curiosity-driven video agent with self-exploration capability, dubbed as VCA. Built upon VLMs, VCA autonomously navigates video segments and efficiently builds a comprehensive understanding of complex video sequences. Instead of directly sampling frames, VCA employs a tree-search structure to explore video segments and collect frames. Rather than relying on external feedback or reward, VCA leverages VLM's self-generated intrinsic reward to guide its exploration, enabling it to capture the most crucial information for reasoning. Experimental results on multiple long video benchmarks demonstrate our approach's superior effectiveness and efficiency.
Recently, with the emergence of recent Multimodal Large Language Model (MLLM) technology, it has become possible to exploit its video understanding capability on different classification tasks. In practice, we face the difficulty of huge requirements for GPU resource if we need to deploy MLLMs online. In this paper, we propose COEF-VQ, a novel cascaded MLLM framework for better video quality understanding on TikTok. To this end, we first propose a MLLM fusing all visual, textual and audio signals, and then develop a cascade framework with a lightweight model as pre-filtering stage and MLLM as fine-consideration stage, significantly reducing the need for GPU resource, while retaining the performance demonstrated solely by MLLM. To demonstrate the effectiveness of COEF-VQ, we deployed this new framework onto the video management platform (VMP) at TikTok, and performed a series of detailed experiments on two in-house tasks related to video quality understanding. We show that COEF-VQ leads to substantial performance gains with limit resource consumption in these two tasks.
This paper presents a comprehensive exploration of the phenomenon of data redundancy in video understanding, with the aim to improve computational efficiency. Our investigation commences with an examination of spatial redundancy, which refers to the observation that the most informative region in each video frame usually corresponds to a small image patch, whose shape, size and location shift smoothly across frames. Motivated by this phenomenon, we formulate the patch localization problem as a dynamic decision task, and introduce a spatially adaptive video recognition approach, termed AdaFocus. In specific, a lightweight encoder is first employed to quickly process the full video sequence, whose features are then utilized by a policy network to identify the most task-relevant regions. Subsequently, the selected patches are inferred by a high-capacity deep network for the final prediction. The full model can be trained in end-to-end conveniently. Furthermore, AdaFocus can be extended by further considering temporal and sample-wise redundancies, i.e., allocating the majority of computation to the most task-relevant frames, and minimizing the computation spent on relatively "easier" videos. Our resulting approach, Uni-AdaFocus, establishes a comprehensive framework that seamlessly integrates spatial, temporal, and sample-wise dynamic computation, while it preserves the merits of AdaFocus in terms of efficient end-to-end training and hardware friendliness. In addition, Uni-AdaFocus is general and flexible as it is compatible with off-the-shelf efficient backbones (e.g., TSM and X3D), which can be readily deployed as our feature extractor, yielding a significantly improved computational efficiency. Empirically, extensive experiments based on seven benchmark datasets and three application scenarios substantiate that Uni-AdaFocus is considerably more efficient than the competitive baselines.
One of the key goals of artificial intelligence (AI) is the development of a multimodal system that facilitates communication with the visual world (image and video) using a natural language query. Earlier works on medical question answering primarily focused on textual and visual (image) modalities, which may be inefficient in answering questions requiring demonstration. In recent years, significant progress has been achieved due to the introduction of large-scale language-vision datasets and the development of efficient deep neural techniques that bridge the gap between language and visual understanding. Improvements have been made in numerous vision-and-language tasks, such as visual captioning visual question answering, and natural language video localization. Most of the existing work on language vision focused on creating datasets and developing solutions for open-domain applications. We believe medical videos may provide the best possible answers to many first aid, medical emergency, and medical education questions. With increasing interest in AI to support clinical decision-making and improve patient engagement, there is a need to explore such challenges and develop efficient algorithms for medical language-video understanding and generation. Toward this, we introduced new tasks to foster research toward designing systems that can understand medical videos to provide visual answers to natural language questions, and are equipped with multimodal capability to generate instruction steps from the medical video. These tasks have the potential to support the development of sophisticated downstream applications that can benefit the public and medical professionals.
Most existing video understanding benchmarks for multimodal large language models (MLLMs) focus only on short videos. The limited number of benchmarks for long video understanding often rely solely on multiple-choice questions (MCQs). However, because of the inherent limitation of MCQ-based evaluation and the increasing reasoning ability of MLLMs, models can give the current answer purely by combining short video understanding with elimination, without genuinely understanding the video content. To address this gap, we introduce CG-Bench, a novel benchmark designed for clue-grounded question answering in long videos. CG-Bench emphasizes the model's ability to retrieve relevant clues for questions, enhancing evaluation credibility. It features 1,219 manually curated videos categorized by a granular system with 14 primary categories, 171 secondary categories, and 638 tertiary categories, making it the largest benchmark for long video analysis. The benchmark includes 12,129 QA pairs in three major question types: perception, reasoning, and hallucination. Compensating the drawbacks of pure MCQ-based evaluation, we design two novel clue-based evaluation methods: clue-grounded white box and black box evaluations, to assess whether the model generates answers based on the correct understanding of the video. We evaluate multiple closed-source and open-source MLLMs on CG-Bench. Results indicate that current models significantly underperform in understanding long videos compared to short ones, and a significant gap exists between open-source and commercial models. We hope CG-Bench can advance the development of more trustworthy and capable MLLMs for long video understanding. All annotations and video data are released at https://cg-bench.github.io/leaderboard/.