{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:W4KZWVU3BFTG2O627PD4U5UYQB","short_pith_number":"pith:W4KZWVU3","schema_version":"1.0","canonical_sha256":"b7159b569b09666d3bdafbc7ca7698806e827b63735beeb568050e9429110725","source":{"kind":"arxiv","id":"2608.03215","version":1},"attestation_state":"computed","paper":{"title":"GROW: Group-Relative Advantage-Weighted On-Policy Reinforcement Learning of Autoregressive-Diffusion Text-to-Speech model","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.CL"],"primary_cat":"eess.AS","authors_text":"Guanrou Yang, Haitao Li, Qian Chen, Qi Chen, Shan Yang, Tian Tan, Wenming Tu, Xie Chen, Yakun Song, Zhikang Niu, Ziyang Ma","submitted_at":"2026-08-04T06:50:55Z","abstract_excerpt":"Reinforcement learning for flow-matching text-to-speech is complicated by deterministic ODE sampling: trajectory-level policy-gradient methods typically convert the ODE into an SDE and track per-step likelihood ratios, introducing stochastic perturbations and substantial overhead. We propose GROW, a group-relative advantage-weighted on-policy RL method that acts directly on the standard flow-matching objective. For each prompt, GROW samples a group of on-policy utterances, separately standardizes intelligibility and speaker-similarity rewards within the group, and combines them to reweight flo"},"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":"2608.03215","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"eess.AS","submitted_at":"2026-08-04T06:50:55Z","cross_cats_sorted":["cs.AI","cs.CL"],"title_canon_sha256":"14c5285d978fcbe3fba461194afaa37bc1de06f4d156680358be16d63cea72d3","abstract_canon_sha256":"75d4d24a3cc1c07223df2873eca6b6b0eaa80539c4a3c317059176ba1a856daa"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-08-05T00:46:17.722112Z","signature_b64":"/CKh4zlLxAKYgkW4YY87Uv8oQiHAyLOBVJRQtQ59c3KUW40K7FLo15TJ8H2J+EO4s6WQFesJfdrOcYfT1tVwBA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"b7159b569b09666d3bdafbc7ca7698806e827b63735beeb568050e9429110725","last_reissued_at":"2026-08-05T00:46:17.719693Z","signature_status":"signed_v1","first_computed_at":"2026-08-05T00:46:17.719693Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"GROW: Group-Relative Advantage-Weighted On-Policy Reinforcement Learning of Autoregressive-Diffusion Text-to-Speech model","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.CL"],"primary_cat":"eess.AS","authors_text":"Guanrou Yang, Haitao Li, Qian Chen, Qi Chen, Shan Yang, Tian Tan, Wenming Tu, Xie Chen, Yakun Song, Zhikang Niu, Ziyang Ma","submitted_at":"2026-08-04T06:50:55Z","abstract_excerpt":"Reinforcement learning for flow-matching text-to-speech is complicated by deterministic ODE sampling: trajectory-level policy-gradient methods typically convert the ODE into an SDE and track per-step likelihood ratios, introducing stochastic perturbations and substantial overhead. We propose GROW, a group-relative advantage-weighted on-policy RL method that acts directly on the standard flow-matching objective. For each prompt, GROW samples a group of on-policy utterances, separately standardizes intelligibility and speaker-similarity rewards within the group, and combines them to reweight flo"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2608.03215","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/2608.03215/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":"2608.03215","created_at":"2026-08-05T00:46:17.720544+00:00"},{"alias_kind":"arxiv_version","alias_value":"2608.03215v1","created_at":"2026-08-05T00:46:17.720544+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2608.03215","created_at":"2026-08-05T00:46:17.720544+00:00"},{"alias_kind":"pith_short_12","alias_value":"W4KZWVU3BFTG","created_at":"2026-08-05T00:46:17.720544+00:00"},{"alias_kind":"pith_short_16","alias_value":"W4KZWVU3BFTG2O62","created_at":"2026-08-05T00:46:17.720544+00:00"},{"alias_kind":"pith_short_8","alias_value":"W4KZWVU3","created_at":"2026-08-05T00:46:17.720544+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/W4KZWVU3BFTG2O627PD4U5UYQB","json":"https://pith.science/pith/W4KZWVU3BFTG2O627PD4U5UYQB.json","graph_json":"https://pith.science/api/pith-number/W4KZWVU3BFTG2O627PD4U5UYQB/graph.json","events_json":"https://pith.science/api/pith-number/W4KZWVU3BFTG2O627PD4U5UYQB/events.json","paper":"https://pith.science/paper/W4KZWVU3"},"agent_actions":{"view_html":"https://pith.science/pith/W4KZWVU3BFTG2O627PD4U5UYQB","download_json":"https://pith.science/pith/W4KZWVU3BFTG2O627PD4U5UYQB.json","view_paper":"https://pith.science/paper/W4KZWVU3","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2608.03215&json=true","fetch_graph":"https://pith.science/api/pith-number/W4KZWVU3BFTG2O627PD4U5UYQB/graph.json","fetch_events":"https://pith.science/api/pith-number/W4KZWVU3BFTG2O627PD4U5UYQB/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/W4KZWVU3BFTG2O627PD4U5UYQB/action/timestamp_anchor","attest_storage":"https://pith.science/pith/W4KZWVU3BFTG2O627PD4U5UYQB/action/storage_attestation","attest_author":"https://pith.science/pith/W4KZWVU3BFTG2O627PD4U5UYQB/action/author_attestation","sign_citation":"https://pith.science/pith/W4KZWVU3BFTG2O627PD4U5UYQB/action/citation_signature","submit_replication":"https://pith.science/pith/W4KZWVU3BFTG2O627PD4U5UYQB/action/replication_record"}},"created_at":"2026-08-05T00:46:17.720544+00:00","updated_at":"2026-08-05T00:46:17.720544+00:00"}