Training-Free Temporal Abstraction for General Video Understanding

• URL: http://arxiv.org/abs/2608.27929v1
• Date: Fri, 28 Aug 2026 05:10:54 GMT
• Title: Training-Free Temporal Abstraction for General Video Understanding
• Abstract summary: We present STITCH, a training-free method that divides a video into semantically meaningful temporal chunks.<n>We evaluate STITCH on generic event boundary detection, language-based moment retrieval, and frame selection for long-video VLM reasoning.
• Score: 12.051446549851844
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Videos are expensive to analyze frame by frame, yet many video understanding tasks depend on knowing where relevant moments occur. A system may need to find when an action changes, locate the segment described by a sentence, or choose a few frames for a vision-language model. Existing methods often solve these problems separately, using task-specific training data or specialized architectures. We study whether a pretrained video-text model can provide enough temporal structure to support several of these tasks at once. We present STITCH, a training-free method that divides a video into semantically meaningful temporal chunks. STITCH embeds short video windows with a frozen video-text backbone and detects changes in the resulting embedding sequence. These chunks are computed once per video and reused across tasks. We evaluate STITCH on generic event boundary detection, language-based moment retrieval, and frame selection for long-video VLM reasoning. Across all three settings, STITCH remains competitive with more specialized methods while requiring no task-specific training, with especially clear gains when only a small number of frames or tokens can be processed. These results suggest that reusable temporal abstraction is a promising direction for general video understanding, allowing dense video streams to be converted once into semantic units that can be localized, retrieved, sampled, or reasoned over by downstream systems.

Related papers

• Improving Spatial-Temporal Reasoning in Video-Language Models with Structured Video Prompting [0.0]
  Video-language models (VLMs) remain brittle on tasks that require tracking events over time and grounding answers in specific spatial regions.<n>We introduce structured video prompting, a training-free inference-time method that augments the input video with lightweight spatial structure and temporal structure.<n>We evaluate this approach on two complementary video benchmarks and two open video-language models.
  arXiv  Detail & Related papers  (2026-08-23T20:16:09Z)
• Multimodal Long Video Modeling Based on Temporal Dynamic Context [13.979661295432964]
  We propose a dynamic long video encoding method utilizing the temporal relationship between frames, named Temporal Dynamic Context (TDC)<n>We segment the video into semantically consistent scenes based on inter-frame similarities, then encode each frame into tokens using visual-audio encoders.<n>To handle extremely long videos, we propose a training-free chain-of-thought strategy that progressively extracts answers from multiple video segments.
  arXiv  Detail & Related papers  (2025-04-14T17:34:06Z)
• Parameter-free Video Segmentation for Vision and Language Understanding [55.20132267309382]
  We propose an algorithm for segmenting videos into contiguous chunks, based on the minimum description length principle.<n>The algorithm is entirely parameter-free, given feature vectors, not requiring a set threshold or the number or size of chunks to be specified.
  arXiv  Detail & Related papers  (2025-03-03T05:54:37Z)
• Disentangling spatio-temporal knowledge for weakly supervised object detection and segmentation in surgical video [10.287675722826028]
  This paper introduces Video Spatio-Temporal Disment Networks (VDST-Net) to disentangle information using semi-decoupled temporal knowledge distillation to predict high-quality class activation maps (CAMs) We demonstrate the efficacy of our framework on a public reference dataset and on a more challenging surgical video dataset where objects are, on average, present in less than 60% of annotated frames.
  arXiv  Detail & Related papers  (2024-07-22T16:52:32Z)
• OmniVid: A Generative Framework for Universal Video Understanding [133.73878582161387]
  We seek to unify the output space of video understanding tasks by using languages as labels and additionally introducing time and box tokens. This enables us to address various types of video tasks, including classification, captioning, and localization. We demonstrate such a simple and straightforward idea is quite effective and can achieve state-of-the-art or competitive results.
  arXiv  Detail & Related papers  (2024-03-26T17:59:24Z)
• Spatio-temporal Prompting Network for Robust Video Feature Extraction [74.54597668310707]
  Frametemporal is one of the main challenges in the field of video understanding. Recent approaches exploit transformer-based integration modules to obtain quality-of-temporal information. We present a neat and unified framework called N-Temporal Prompting Network (NNSTP) It can efficiently extract video features by adjusting the input features in the network backbone.
  arXiv  Detail & Related papers  (2024-02-04T17:52:04Z)
• RTQ: Rethinking Video-language Understanding Based on Image-text Model [55.278942477715084]
  Video-language understanding presents unique challenges due to the inclusion of highly complex semantic details. We propose a novel framework called RTQ, which addresses these challenges simultaneously. Our model demonstrates outstanding performance even in the absence of video-language pre-training.
  arXiv  Detail & Related papers  (2023-12-01T04:51:01Z)
• Temporal Perceiving Video-Language Pre-training [112.1790287726804]
  This work introduces a novel text-video localization pre-text task to enable fine-grained temporal and semantic alignment. Specifically, text-video localization consists of moment retrieval, which predicts start and end boundaries in videos given the text description. Our method connects the fine-grained frame representations with the word representations and implicitly distinguishes representations of different instances in the single modality.
  arXiv  Detail & Related papers  (2023-01-18T12:15:47Z)
• Revealing Single Frame Bias for Video-and-Language Learning [115.01000652123882]
  We show that a single-frame trained model can achieve better performance than existing methods that use multiple frames for training. This result reveals the existence of a strong "static appearance bias" in popular video-and-language datasets. We propose two new retrieval tasks based on existing fine-grained action recognition datasets that encourage temporal modeling.
  arXiv  Detail & Related papers  (2022-06-07T16:28:30Z)
• A Hierarchical Multi-Modal Encoder for Moment Localization in Video Corpus [31.387948069111893]
  We show how to identify a short segment in a long video that semantically matches a text query. To tackle this problem, we propose the HierArchical Multi-Modal EncodeR (HAMMER) that encodes a video at both the coarse-grained clip level and the fine-trimmed frame level. We conduct extensive experiments to evaluate our model on moment localization in video corpus on ActivityNet Captions and TVR datasets.
  arXiv  Detail & Related papers  (2020-11-18T02:42:36Z)

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.