MusAV is a new public benchmark dataset for comparative validation of arousal and valence (AV) regression models for audio-based music emotion recognition. We built MusAV by gathering comparative annotations of arousal and valence on pairs of music tracks, using track audio previews and metadata from the Spotify API. The resulting dataset contains 2,092 track previews covering 1,404 genres, with pairwise relative AV judgments by 20 annotators and various subsets of the ground truth based on different levels of annotation agreement. This repository contains the dataset metadata, audio track previews and metadata gathered from the Spotify API for the annotated chunks. Please see the companion website and the related GitHub repository for more information on how to use the dataset and evaluation scripts. Citation If you use the MusAV Dataset, please cite our ISMIR 2022 paper: Bogdanov, D., Lizarraga-Seijas, X., Alonso-Jiménez, P., & Serra X. (2022). MusAV: A dataset of relative arousal-valence annotations for validation of audio models. International Society for Music Information Retrieval Conference (ISMIR 2022). BibTeX: @conference {bogdanov2019mtg, author = "Bogdanov, Dmitry and Lizarraga-Seijas, Xavier and Alonso-Jiménez, Pablo and Serra, Xavier", title = "MusAV: A dataset of relative arousal-valence annotations for validation of audio models", booktitle = "International Society for Music Information Retrieval Conference (ISMIR 2022)", year = "2022", address = "Bengaluru, India", url = "http://hdl.handle.net/10230/54181" } Acknowledgments This research was carried out under the project Musical AI - PID2019-111403GB-I00/AEI/10.13039/501100011033, funded by the Spanish Ministerio de Ciencia e Innovación and the Agencia Estatal de Investigación. We thank all the annotators who participated in the creation of the dataset.