{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:ZJYIJ3HRMJST5JZPMG2AVSUHXS","short_pith_number":"pith:ZJYIJ3HR","schema_version":"1.0","canonical_sha256":"ca7084ecf162653ea72f61b40aca87bc8007b08e6d370464e9171e345cf3234c","source":{"kind":"arxiv","id":"2605.22083","version":1},"attestation_state":"computed","paper":{"title":"RobustSpeechFlow: Learning Robust Text-to-Speech Trajectories via Augmentation-based Contrastive Flow Matching","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG","eess.AS"],"primary_cat":"cs.SD","authors_text":"Frederik Bous, Hyeongju Kim, Jinhyeok Yang, Joon Byun, Juheon Lee, Yechan Yu","submitted_at":"2026-05-21T07:22:28Z","abstract_excerpt":"While flow-matching text-to-speech (TTS) achieves strong zero-shot speaker similarity and naturalness, it remains susceptible to content fidelity issues, particularly skip and repeat errors from imperfect alignment. We propose RobustSpeechFlow, a training strategy that improves alignment robustness by extending contrastive flow matching with length-preserving repeat and skip latent augmentations. Requiring no external aligners or preference data, our method directly penalizes realistic failure modes and readily integrates into existing pipelines. On Seed-TTS-eval, it reduces the word error rat"},"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":"2605.22083","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.SD","submitted_at":"2026-05-21T07:22:28Z","cross_cats_sorted":["cs.LG","eess.AS"],"title_canon_sha256":"92baeea7bb64cfef676b459a2d94dfa881d1a7d7f147e86a6ccbf9c3a265df3d","abstract_canon_sha256":"67a2a88b42d3d61a71df1130a913fd259e3364050750b0c41cd280e59a694396"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-22T01:04:24.545450Z","signature_b64":"l1iiLLP+D6rY36HdqRk+kp4yU7kzfPo4FIjIyw1WuedbJ2ogXPJaUw3nGNAJVS/t+pvtwKCsyqkvkKcTry5NCw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"ca7084ecf162653ea72f61b40aca87bc8007b08e6d370464e9171e345cf3234c","last_reissued_at":"2026-05-22T01:04:24.544494Z","signature_status":"signed_v1","first_computed_at":"2026-05-22T01:04:24.544494Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"RobustSpeechFlow: Learning Robust Text-to-Speech Trajectories via Augmentation-based Contrastive Flow Matching","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG","eess.AS"],"primary_cat":"cs.SD","authors_text":"Frederik Bous, Hyeongju Kim, Jinhyeok Yang, Joon Byun, Juheon Lee, Yechan Yu","submitted_at":"2026-05-21T07:22:28Z","abstract_excerpt":"While flow-matching text-to-speech (TTS) achieves strong zero-shot speaker similarity and naturalness, it remains susceptible to content fidelity issues, particularly skip and repeat errors from imperfect alignment. We propose RobustSpeechFlow, a training strategy that improves alignment robustness by extending contrastive flow matching with length-preserving repeat and skip latent augmentations. Requiring no external aligners or preference data, our method directly penalizes realistic failure modes and readily integrates into existing pipelines. On Seed-TTS-eval, it reduces the word error rat"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.22083","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/2605.22083/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":"2605.22083","created_at":"2026-05-22T01:04:24.544663+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.22083v1","created_at":"2026-05-22T01:04:24.544663+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.22083","created_at":"2026-05-22T01:04:24.544663+00:00"},{"alias_kind":"pith_short_12","alias_value":"ZJYIJ3HRMJST","created_at":"2026-05-22T01:04:24.544663+00:00"},{"alias_kind":"pith_short_16","alias_value":"ZJYIJ3HRMJST5JZP","created_at":"2026-05-22T01:04:24.544663+00:00"},{"alias_kind":"pith_short_8","alias_value":"ZJYIJ3HR","created_at":"2026-05-22T01:04:24.544663+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2605.22083","citing_title":"RobustSpeechFlow: Learning Robust Text-to-Speech Trajectories via Augmentation-based Contrastive Flow Matching","ref_index":1,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/ZJYIJ3HRMJST5JZPMG2AVSUHXS","json":"https://pith.science/pith/ZJYIJ3HRMJST5JZPMG2AVSUHXS.json","graph_json":"https://pith.science/api/pith-number/ZJYIJ3HRMJST5JZPMG2AVSUHXS/graph.json","events_json":"https://pith.science/api/pith-number/ZJYIJ3HRMJST5JZPMG2AVSUHXS/events.json","paper":"https://pith.science/paper/ZJYIJ3HR"},"agent_actions":{"view_html":"https://pith.science/pith/ZJYIJ3HRMJST5JZPMG2AVSUHXS","download_json":"https://pith.science/pith/ZJYIJ3HRMJST5JZPMG2AVSUHXS.json","view_paper":"https://pith.science/paper/ZJYIJ3HR","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.22083&json=true","fetch_graph":"https://pith.science/api/pith-number/ZJYIJ3HRMJST5JZPMG2AVSUHXS/graph.json","fetch_events":"https://pith.science/api/pith-number/ZJYIJ3HRMJST5JZPMG2AVSUHXS/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/ZJYIJ3HRMJST5JZPMG2AVSUHXS/action/timestamp_anchor","attest_storage":"https://pith.science/pith/ZJYIJ3HRMJST5JZPMG2AVSUHXS/action/storage_attestation","attest_author":"https://pith.science/pith/ZJYIJ3HRMJST5JZPMG2AVSUHXS/action/author_attestation","sign_citation":"https://pith.science/pith/ZJYIJ3HRMJST5JZPMG2AVSUHXS/action/citation_signature","submit_replication":"https://pith.science/pith/ZJYIJ3HRMJST5JZPMG2AVSUHXS/action/replication_record"}},"created_at":"2026-05-22T01:04:24.544663+00:00","updated_at":"2026-05-22T01:04:24.544663+00:00"}