{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:HR4NTQBPARASWURHRPX5BANW3A","short_pith_number":"pith:HR4NTQBP","schema_version":"1.0","canonical_sha256":"3c78d9c02f04412b52278befd081b6d82aebf2c7fb9b2050823c61c91c715755","source":{"kind":"arxiv","id":"2402.04229","version":1},"attestation_state":"computed","paper":{"title":"MusicRL: Aligning Music Generation to Human Preferences","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.SD","eess.AS"],"primary_cat":"cs.LG","authors_text":"Andrea Agostinelli, Brian McWilliams, Damien Vincent, Geoffrey Cideron, L\\'eonard Hussenot, Matej Kastelic, Matthieu Geist, Mauro Verzetti, Neil Zeghidour, Olivier Bachem, Olivier Pietquin, Sertan Girgin, Victor Ungureanu, Zal\\'an Borsos","submitted_at":"2024-02-06T18:36:52Z","abstract_excerpt":"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"},"verification_status":{"content_addressed":true,"pith_receipt":true,"author_attested":false,"weak_author_claims":0,"strong_author_claims":0,"externally_anchored":false,"storage_verified":false,"citation_signatures":0,"replication_records":0,"graph_snapshot":true,"references_resolved":false,"formal_links_present":false},"canonical_record":{"source":{"id":"2402.04229","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-02-06T18:36:52Z","cross_cats_sorted":["cs.SD","eess.AS"],"title_canon_sha256":"cf329a95e8fd3f72643c4a68e33d812e3b6dd526ae9792e216bbb2d7599344c2","abstract_canon_sha256":"c51bc17b9a422ee09020c586e03e37bbdaa39f8d923ce61e304601f475b1eaa0"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T07:42:05.220658Z","signature_b64":"r+0+FfhDcv+9a0P+Y8wm14WFT4GrxIe9RP5n/j6mpsGEjQDIACIp79+pPMxalBL2jN6mPYwI4c3hhMCd3x5oBA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"3c78d9c02f04412b52278befd081b6d82aebf2c7fb9b2050823c61c91c715755","last_reissued_at":"2026-07-05T07:42:05.220149Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T07:42:05.220149Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"MusicRL: Aligning Music Generation to Human Preferences","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.SD","eess.AS"],"primary_cat":"cs.LG","authors_text":"Andrea Agostinelli, Brian McWilliams, Damien Vincent, Geoffrey Cideron, L\\'eonard Hussenot, Matej Kastelic, Matthieu Geist, Mauro Verzetti, Neil Zeghidour, Olivier Bachem, Olivier Pietquin, Sertan Girgin, Victor Ungureanu, Zal\\'an Borsos","submitted_at":"2024-02-06T18:36:52Z","abstract_excerpt":"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"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2402.04229","kind":"arxiv","version":1},"verdict":{"id":null,"model_set":{},"created_at":null,"strongest_claim":"","one_line_summary":"","pipeline_version":null,"weakest_assumption":"","pith_extraction_headline":""},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2402.04229/integrity.json","findings":[],"available":true,"detectors_run":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"aliases":[{"alias_kind":"arxiv","alias_value":"2402.04229","created_at":"2026-07-05T07:42:05.220207+00:00"},{"alias_kind":"arxiv_version","alias_value":"2402.04229v1","created_at":"2026-07-05T07:42:05.220207+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2402.04229","created_at":"2026-07-05T07:42:05.220207+00:00"},{"alias_kind":"pith_short_12","alias_value":"HR4NTQBPARAS","created_at":"2026-07-05T07:42:05.220207+00:00"},{"alias_kind":"pith_short_16","alias_value":"HR4NTQBPARASWURH","created_at":"2026-07-05T07:42:05.220207+00:00"},{"alias_kind":"pith_short_8","alias_value":"HR4NTQBP","created_at":"2026-07-05T07:42:05.220207+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2406.02430","citing_title":"Seed-TTS: A Family of High-Quality Versatile Speech Generation Models","ref_index":39,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/HR4NTQBPARASWURHRPX5BANW3A","json":"https://pith.science/pith/HR4NTQBPARASWURHRPX5BANW3A.json","graph_json":"https://pith.science/api/pith-number/HR4NTQBPARASWURHRPX5BANW3A/graph.json","events_json":"https://pith.science/api/pith-number/HR4NTQBPARASWURHRPX5BANW3A/events.json","paper":"https://pith.science/paper/HR4NTQBP"},"agent_actions":{"view_html":"https://pith.science/pith/HR4NTQBPARASWURHRPX5BANW3A","download_json":"https://pith.science/pith/HR4NTQBPARASWURHRPX5BANW3A.json","view_paper":"https://pith.science/paper/HR4NTQBP","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2402.04229&json=true","fetch_graph":"https://pith.science/api/pith-number/HR4NTQBPARASWURHRPX5BANW3A/graph.json","fetch_events":"https://pith.science/api/pith-number/HR4NTQBPARASWURHRPX5BANW3A/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/HR4NTQBPARASWURHRPX5BANW3A/action/timestamp_anchor","attest_storage":"https://pith.science/pith/HR4NTQBPARASWURHRPX5BANW3A/action/storage_attestation","attest_author":"https://pith.science/pith/HR4NTQBPARASWURHRPX5BANW3A/action/author_attestation","sign_citation":"https://pith.science/pith/HR4NTQBPARASWURHRPX5BANW3A/action/citation_signature","submit_replication":"https://pith.science/pith/HR4NTQBPARASWURHRPX5BANW3A/action/replication_record"}},"created_at":"2026-07-05T07:42:05.220207+00:00","updated_at":"2026-07-05T07:42:05.220207+00:00"}