TimeSearch: Hierarchical Video Search with Spotlight and Reflection for Human-like Long Video Understanding

• URL: http://arxiv.org/abs/2504.01407v1
• Date: Wed, 02 Apr 2025 06:47:19 GMT
• Title: TimeSearch: Hierarchical Video Search with Spotlight and Reflection for Human-like Long Video Understanding
• Authors: Junwen Pan, Rui Zhang, Xin Wan, Yuan Zhang, Ming Lu, Qi She,
• Abstract summary: Large video-language models (LVLMs) have shown remarkable performance across various video-language tasks.<n>Downsampling long videos in either space or time can lead to visual hallucinations, making it difficult to accurately interpret long videos.<n>TimeSearch integrates two human-like primitives into a unified autoregressive LVLM.
• Score: 24.52604124233087
• License: http://creativecommons.org/licenses/by-nc-sa/4.0/
• Abstract: Large video-language models (LVLMs) have shown remarkable performance across various video-language tasks. However, they encounter significant challenges when processing long videos because of the large number of video frames involved. Downsampling long videos in either space or time can lead to visual hallucinations, making it difficult to accurately interpret long videos. Motivated by human hierarchical temporal search strategies, we propose \textbf{TimeSearch}, a novel framework enabling LVLMs to understand long videos in a human-like manner. TimeSearch integrates two human-like primitives into a unified autoregressive LVLM: 1) \textbf{Spotlight} efficiently identifies relevant temporal events through a Temporal-Augmented Frame Representation (TAFR), explicitly binding visual features with timestamps; 2) \textbf{Reflection} evaluates the correctness of the identified events, leveraging the inherent temporal self-reflection capabilities of LVLMs. TimeSearch progressively explores key events and prioritizes temporal search based on reflection confidence. Extensive experiments on challenging long-video benchmarks confirm that TimeSearch substantially surpasses previous state-of-the-art, improving the accuracy from 41.8\% to 51.5\% on the LVBench. Additionally, experiments on temporal grounding demonstrate that appropriate TAFR is adequate to effectively stimulate the surprising temporal grounding ability of LVLMs in a simpler yet versatile manner, which improves mIoU on Charades-STA by 11.8\%. The code will be released.

Related papers

• Re-thinking Temporal Search for Long-Form Video Understanding [67.12801626407135]
  Current temporal search methods only achieve 2.1% temporal F1 score on the Longvideobench subset. Inspired by visual search in images, we propose a lightweight temporal search framework, T* that reframes costly temporal search as spatial search. Extensive experiments show that integrating T* with existing methods significantly improves SOTA long-form video understanding.
  arXiv  Detail & Related papers  (2025-04-03T04:03:10Z)
• Token-Efficient Long Video Understanding for Multimodal LLMs [101.70681093383365]
  STORM is a novel architecture incorporating a dedicated temporal encoder between the image encoder and the Video-LLMs.<n>We show that STORM achieves state-of-the-art results across various long video understanding benchmarks.
  arXiv  Detail & Related papers  (2025-03-06T06:17:38Z)
• MomentSeeker: A Comprehensive Benchmark and A Strong Baseline For Moment Retrieval Within Long Videos [62.01402470874109]
  We present MomentSeeker, a benchmark to evaluate retrieval models' performance in handling general long-video moment retrieval tasks.<n>It incorporates long videos of over 500 seconds on average, making it the first benchmark specialized for long-video moment retrieval.<n>It covers a wide range of task categories (including Moment Search, Caption Alignment, Image-conditioned Moment Search, and Video-conditioned Moment Search) and diverse application scenarios.<n>We further fine-tune an MLLM-based LVMR retriever on synthetic data, which demonstrates strong performance on our benchmark.
  arXiv  Detail & Related papers  (2025-02-18T05:50:23Z)
• CaReBench: A Fine-Grained Benchmark for Video Captioning and Retrieval [24.203328970223527]
  We present CaReBench, a testing benchmark for fine-grained video captioning and retrieval.<n>Uniquely, it provides manually separated spatial annotations and temporal annotations for each video.<n>Based on this design, we introduce two evaluation metrics, ReBias and CapST, specifically tailored for video retrieval and video captioning tasks.
  arXiv  Detail & Related papers  (2024-12-31T15:53:50Z)
• Reversed in Time: A Novel Temporal-Emphasized Benchmark for Cross-Modal Video-Text Retrieval [56.05621657583251]
  Cross-modal (e.g. image-text, video-text) retrieval is an important task in information retrieval and multimodal vision-language understanding field.<n>We introduce RTime, a novel temporal-emphasized video-text retrieval dataset.<n>Our RTime dataset currently consists of 21k videos with 10 captions per video, totalling about 122 hours.
  arXiv  Detail & Related papers  (2024-12-26T11:32:00Z)
• Temporal Reasoning Transfer from Text to Video [51.68487044397409]
  Video Large Language Models (Video LLMs) struggle with tracking temporal changes and reasoning about temporal relationships. We introduce the Textual Temporal reasoning Transfer (T3) to transfer temporal reasoning abilities from text to video domains. LongVA-7B model achieves competitive performance on comprehensive video benchmarks.
  arXiv  Detail & Related papers  (2024-10-08T16:10:29Z)
• T2VIndexer: A Generative Video Indexer for Efficient Text-Video Retrieval [30.48217069475297]
  We introduce a model-based video indexer named T2VIndexer, which is a sequence-to-sequence generative model directly generating video identifiers. T2VIndexer aims to reduce retrieval time while maintaining high accuracy.
  arXiv  Detail & Related papers  (2024-08-21T08:40:45Z)
• LITA: Language Instructed Temporal-Localization Assistant [71.68815100776278]
  We introduce time tokens that encode timestamps relative to the video length to better represent time in videos. We also introduce SlowFast tokens in the architecture to capture temporal information at fine temporal resolution. We show that our emphasis on temporal localization also substantially improves video-based text generation compared to existing Video LLMs.
  arXiv  Detail & Related papers  (2024-03-27T22:50:48Z)
• VTimeLLM: Empower LLM to Grasp Video Moments [43.51980030572101]
  Large language models (LLMs) have shown remarkable text understanding capabilities. Video LLMs can only provide a coarse description of the entire video. We propose VTimeLLM, a novel Video LLM for fine-grained video moment understanding.
  arXiv  Detail & Related papers  (2023-11-30T10:49:56Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.