TextVidBench: A Benchmark for Long Video Scene Text Understanding

TL;DR

TextVidBench is a comprehensive benchmark for evaluating long-video scene text understanding, addressing limitations of previous datasets by covering diverse domains, providing detailed annotations, and proposing methods to enhance model performance on videos over three minutes long.

Contribution

The paper introduces TextVidBench, the first long-video scene text understanding benchmark with diverse domain coverage, a three-stage evaluation framework, and high-quality annotations, along with novel methods to improve model performance.

Findings

Existing models struggle with long-video scene text understanding.

The proposed methods improve temporal perception in large models.

TextVidBench presents significant challenges for current models.

Abstract

Despite recent progress on the short-video Text-Visual Question Answering (ViteVQA) task - largely driven by benchmarks such as M4-ViteVQA - existing datasets still suffer from limited video duration and narrow evaluation scopes, making it difficult to adequately assess the growing capabilities of powerful multimodal large language models (MLLMs). To address these limitations, we introduce TextVidBench, the first benchmark specifically designed for long-video text question answering (>3 minutes). TextVidBench makes three key contributions: 1) Cross-domain long-video coverage: Spanning 9 categories (e.g., news, sports, gaming), with an average video length of 2306 seconds, enabling more realistic evaluation of long-video understanding. 2) A three-stage evaluation framework: "Text Needle-in-Haystack -> Temporal Grounding -> Text Dynamics Captioning". 3) High-quality fine-grained annotations: Containing over 5,000 question-answer pairs with detailed semantic labeling. Furthermore, we propose an efficient paradigm for improving large models through: (i) introducing the IT-Rope mechanism and temporal prompt engineering to enhance temporal perception, (ii) adopting non-uniform positional encoding to better handle long video sequences, and (iii) applying lightweight fine-tuning on video-text data. Extensive experiments on multiple public datasets as well as TextVidBench demonstrate that our new benchmark presents significant challenges to existing models, while our proposed method offers valuable insights into improving long-video scene text understanding capabilities.