![[2602.17814] VQPP: Video Query Performance Prediction Benchmark](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2602.17814] VQPP: Video Query Performance Prediction Benchmark
Summary
The paper introduces the Video Query Performance Prediction (VQPP) benchmark, addressing a gap in query performance prediction for video retrieval. It includes datasets and systems for evaluating performance predictors in content-based video retrieval.
Why It Matters
As video content becomes increasingly prevalent, effective query performance prediction is crucial for enhancing retrieval systems. This benchmark enables researchers to explore and improve video query performance, fostering advancements in information retrieval and machine learning applications.
Key Takeaways
- VQPP is the first benchmark for video query performance prediction.
- It includes two datasets and two systems for comprehensive evaluation.
- Pre-retrieval predictors show competitive performance, aiding in retrieval processes.
- The benchmark supports reproducible results and direct comparisons.
- The best performing predictor can be utilized in training large language models.
Computer Science > Computer Vision and Pattern Recognition arXiv:2602.17814 (cs) [Submitted on 19 Feb 2026] Title:VQPP: Video Query Performance Prediction Benchmark Authors:Adrian Catalin Lutu, Eduard Poesina, Radu Tudor Ionescu View a PDF of the paper titled VQPP: Video Query Performance Prediction Benchmark, by Adrian Catalin Lutu and 2 other authors View PDF HTML (experimental) Abstract:Query performance prediction (QPP) is an important and actively studied information retrieval task, having various applications, such as query reformulation, query expansion, and retrieval system selection, among many others. The task has been primarily studied in the context of text and image retrieval, whereas QPP for content-based video retrieval (CBVR) remains largely underexplored. To this end, we propose the first benchmark for video query performance prediction (VQPP), comprising two text-to-video retrieval datasets and two CBVR systems, respectively. VQPP contains a total of 56K text queries and 51K videos, and comes with official training, validation and test splits, fostering direct comparisons and reproducible results. We explore multiple pre-retrieval and post-retrieval performance predictors, creating a representative benchmark for future exploration of QPP in the video domain. Our results show that pre-retrieval predictors obtain competitive performance, enabling applications before performing the retrieval step. We also demonstrate the applicability of VQPP by employing th...