MusicRL: Aligning Music Generation to Human Preferences
Fuente:
arXiv
Salvato in:
| Autori principali: | , , , , , , , , , , , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2024
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866916117183725568 |
|---|---|
| author | Cideron, Geoffrey Girgin, Sertan Verzetti, Mauro Vincent, Damien Kastelic, Matej Borsos, Zalán McWilliams, Brian Ungureanu, Victor Bachem, Olivier Pietquin, Olivier Geist, Matthieu Hussenot, Léonard Zeghidour, Neil Agostinelli, Andrea |
| author_facet | Cideron, Geoffrey Girgin, Sertan Verzetti, Mauro Vincent, Damien Kastelic, Matej Borsos, Zalán McWilliams, Brian Ungureanu, Victor Bachem, Olivier Pietquin, Olivier Geist, Matthieu Hussenot, Léonard Zeghidour, Neil Agostinelli, Andrea |
| contents | We propose MusicRL, the first music generation system finetuned from human feedback. Appreciation of text-to-music models is particularly subjective since the concept of musicality as well as the specific intention behind a caption are user-dependent (e.g. a caption such as "upbeat work-out music" can map to a retro guitar solo or a techno pop beat). Not only this makes supervised training of such models challenging, but it also calls for integrating continuous human feedback in their post-deployment finetuning. MusicRL is a pretrained autoregressive MusicLM (Agostinelli et al., 2023) model of discrete audio tokens finetuned with reinforcement learning to maximise sequence-level rewards. We design reward functions related specifically to text-adherence and audio quality with the help from selected raters, and use those to finetune MusicLM into MusicRL-R. We deploy MusicLM to users and collect a substantial dataset comprising 300,000 pairwise preferences. Using Reinforcement Learning from Human Feedback (RLHF), we train MusicRL-U, the first text-to-music model that incorporates human feedback at scale. Human evaluations show that both MusicRL-R and MusicRL-U are preferred to the baseline. Ultimately, MusicRL-RU combines the two approaches and results in the best model according to human raters. Ablation studies shed light on the musical attributes influencing human preferences, indicating that text adherence and quality only account for a part of it. This underscores the prevalence of subjectivity in musical appreciation and calls for further involvement of human listeners in the finetuning of music generation models. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2402_04229 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | MusicRL: Aligning Music Generation to Human Preferences Cideron, Geoffrey Girgin, Sertan Verzetti, Mauro Vincent, Damien Kastelic, Matej Borsos, Zalán McWilliams, Brian Ungureanu, Victor Bachem, Olivier Pietquin, Olivier Geist, Matthieu Hussenot, Léonard Zeghidour, Neil Agostinelli, Andrea Machine Learning Sound Audio and Speech Processing We propose MusicRL, the first music generation system finetuned from human feedback. Appreciation of text-to-music models is particularly subjective since the concept of musicality as well as the specific intention behind a caption are user-dependent (e.g. a caption such as "upbeat work-out music" can map to a retro guitar solo or a techno pop beat). Not only this makes supervised training of such models challenging, but it also calls for integrating continuous human feedback in their post-deployment finetuning. MusicRL is a pretrained autoregressive MusicLM (Agostinelli et al., 2023) model of discrete audio tokens finetuned with reinforcement learning to maximise sequence-level rewards. We design reward functions related specifically to text-adherence and audio quality with the help from selected raters, and use those to finetune MusicLM into MusicRL-R. We deploy MusicLM to users and collect a substantial dataset comprising 300,000 pairwise preferences. Using Reinforcement Learning from Human Feedback (RLHF), we train MusicRL-U, the first text-to-music model that incorporates human feedback at scale. Human evaluations show that both MusicRL-R and MusicRL-U are preferred to the baseline. Ultimately, MusicRL-RU combines the two approaches and results in the best model according to human raters. Ablation studies shed light on the musical attributes influencing human preferences, indicating that text adherence and quality only account for a part of it. This underscores the prevalence of subjectivity in musical appreciation and calls for further involvement of human listeners in the finetuning of music generation models. |
| title | MusicRL: Aligning Music Generation to Human Preferences |
| topic | Machine Learning Sound Audio and Speech Processing |
| url | https://arxiv.org/abs/2402.04229 |