---
title: "SCVBench: A Benchmark with Multi-turn Dialogues for Story-Centric Video Understanding"
authors:
  - "Sisi You"
  - "Bowen Yuan"
  - "Bing-Kun Bao"
doi: "10.24963/ijcai.2025/255"
url: "https://ijcai.org/proceedings/2025/255"
pdf: "https://ijcai.org/proceedings/2025/0255.pdf"
venue: "Proceedings of the Thirty-Fourth International Joint Conference on Artificial Intelligence"
year: 2025
pages: "2287-2295"
track: "Main Track"
keywords:
  - "Computer Vision: CV: Multimodal learning"
  - "Computer Vision: CV: Vision, language and reasoning"
  - "Computer Vision: CV: Video analysis and understanding"
---

# SCVBench: A Benchmark with Multi-turn Dialogues for Story-Centric Video Understanding

Sisi You, Bowen Yuan, Bing-Kun Bao

*Proceedings of the Thirty-Fourth International Joint Conference on Artificial Intelligence*, pages 2287-2295. DOI: [10.24963/ijcai.2025/255](https://doi.org/10.24963/ijcai.2025/255)

## Abstract

Video understanding seeks to enable machines to interpret visual content across three levels: action, event, and story.  Existing models are limited in their ability to perform high-level long-term story understanding, due to (1) the oversimplified treatment of temporal information and (2) the training bias introduced by action/event-centric datasets. To address this, we introduce SCVBench, a novel benchmark for story-centric video understanding. SCVBench evaluates LVLMs through an event ordering task decomposed into sub-questions leading to a final question, quantitatively measuring historical dialogue exploration. We collected 1,253 final questions and 6,027 sub-question pairs from 925 videos, constructing continuous multi-turn dialogues. Experimental results show that while closed-source GPT-4o outperforms other models, most open-source LVLMs struggle with story-centric video understanding. Additionally, our StoryCoT model significantly surpasses open-source LVLMs on SCVBench. SCVBench aims to advance research by comprehensively analyzing LVLMs' temporal reasoning and comprehension capabilities. Code can be accessed at https://github.com/yuanrr/SCVBench.