Summary of Evaluation Of Pretrained Language Models on Music Understanding, by Yannis Vasilakis et al.
Evaluation of pretrained language models on music understanding
by Yannis Vasilakis, Rachel Bittner, Johan Pauwels
First submitted to arxiv on: 17 Sep 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Artificial Intelligence (cs.AI); Information Retrieval (cs.IR); Sound (cs.SD); Audio and Speech Processing (eess.AS)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary The paper investigates the musical knowledge of Large Language Models (LLMs) in Music Information Research (MIR) applications. Despite their reported success, LLMs are found to suffer from prompt sensitivity, inability to model negation, and sensitivity towards specific words. The authors quantify these properties using triplet-based accuracy, evaluating the ability to model relative similarity of labels in a hierarchical ontology. The Audioset ontology is leveraged to generate triplets for genre and instruments sub-trees. Six general-purpose Transformer-based models are evaluated, revealing inconsistencies in all six models, suggesting that off-the-shelf LLMs need adaptation to music before use. Keywords: Music Information Research, Large Language Models, transformer-based models, Audioset ontology. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper looks at how well computers can understand music. Despite their success, these computer models are found to have some limitations when it comes to understanding music. They struggle with things like understanding what makes a song “rock” or not having guitars, and they’re sensitive to certain words. The researchers use a special way of testing called triplets to see how well the computers do at understanding music. They find that all six computer models have some flaws, which means they need to be changed before they can be used for music tasks. |
Keywords
» Artificial intelligence » Prompt » Transformer
