{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:DBG5FU5Z7IA6ELRFXCGUDKM4PA","short_pith_number":"pith:DBG5FU5Z","schema_version":"1.0","canonical_sha256":"184dd2d3b9fa01e22e25b88d41a99c7832d29f57a89196285e4b975b7357708f","source":{"kind":"arxiv","id":"2406.17957","version":1},"attestation_state":"computed","paper":{"title":"Improving Robustness of LLM-based Speech Synthesis by Learning Monotonic Alignment","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","eess.AS"],"primary_cat":"cs.SD","authors_text":"Boris Ginsburg, Jason Li, Paarth Neekhara, Rafael Valle, Rohan Badlani, Shehzeen Hussain, Subhankar Ghosh","submitted_at":"2024-06-25T22:18:52Z","abstract_excerpt":"Large Language Model (LLM) based text-to-speech (TTS) systems have demonstrated remarkable capabilities in handling large speech datasets and generating natural speech for new speakers. However, LLM-based TTS models are not robust as the generated output can contain repeating words, missing words and mis-aligned speech (referred to as hallucinations or attention errors), especially when the text contains multiple occurrences of the same token. We examine these challenges in an encoder-decoder transformer model and find that certain cross-attention heads in such models implicitly learn the text"},"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":"2406.17957","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.SD","submitted_at":"2024-06-25T22:18:52Z","cross_cats_sorted":["cs.AI","eess.AS"],"title_canon_sha256":"9c3442af7944ed77ccb89bfe846eb15c353243432391fa2e6e8afd2135486ca1","abstract_canon_sha256":"7ac3f2ebb47578d010abcd8f8dcf317eb9b6bacd9667460655d5449e6b80884e"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:36:52.929125Z","signature_b64":"90/467aX8Rj2EgE6rDN5sjtYNE425UsHPg72zpL6x/m34s1UxGfgbXjWwPCe7dwzskkufFpWvZX+s3hezdATBw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"184dd2d3b9fa01e22e25b88d41a99c7832d29f57a89196285e4b975b7357708f","last_reissued_at":"2026-07-05T08:36:52.928650Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:36:52.928650Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Improving Robustness of LLM-based Speech Synthesis by Learning Monotonic Alignment","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","eess.AS"],"primary_cat":"cs.SD","authors_text":"Boris Ginsburg, Jason Li, Paarth Neekhara, Rafael Valle, Rohan Badlani, Shehzeen Hussain, Subhankar Ghosh","submitted_at":"2024-06-25T22:18:52Z","abstract_excerpt":"Large Language Model (LLM) based text-to-speech (TTS) systems have demonstrated remarkable capabilities in handling large speech datasets and generating natural speech for new speakers. However, LLM-based TTS models are not robust as the generated output can contain repeating words, missing words and mis-aligned speech (referred to as hallucinations or attention errors), especially when the text contains multiple occurrences of the same token. We examine these challenges in an encoder-decoder transformer model and find that certain cross-attention heads in such models implicitly learn the text"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2406.17957","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/2406.17957/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":"2406.17957","created_at":"2026-07-05T08:36:52.928708+00:00"},{"alias_kind":"arxiv_version","alias_value":"2406.17957v1","created_at":"2026-07-05T08:36:52.928708+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2406.17957","created_at":"2026-07-05T08:36:52.928708+00:00"},{"alias_kind":"pith_short_12","alias_value":"DBG5FU5Z7IA6","created_at":"2026-07-05T08:36:52.928708+00:00"},{"alias_kind":"pith_short_16","alias_value":"DBG5FU5Z7IA6ELRF","created_at":"2026-07-05T08:36:52.928708+00:00"},{"alias_kind":"pith_short_8","alias_value":"DBG5FU5Z","created_at":"2026-07-05T08:36:52.928708+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.06582","citing_title":"PairAlign: A Framework for Sequence Tokenization via Self-Alignment with Applications to Audio Tokenization","ref_index":38,"is_internal_anchor":false},{"citing_arxiv_id":"2605.06582","citing_title":"PairAlign: A Framework for Sequence Tokenization via Self-Alignment with Applications to Audio Tokenization","ref_index":38,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/DBG5FU5Z7IA6ELRFXCGUDKM4PA","json":"https://pith.science/pith/DBG5FU5Z7IA6ELRFXCGUDKM4PA.json","graph_json":"https://pith.science/api/pith-number/DBG5FU5Z7IA6ELRFXCGUDKM4PA/graph.json","events_json":"https://pith.science/api/pith-number/DBG5FU5Z7IA6ELRFXCGUDKM4PA/events.json","paper":"https://pith.science/paper/DBG5FU5Z"},"agent_actions":{"view_html":"https://pith.science/pith/DBG5FU5Z7IA6ELRFXCGUDKM4PA","download_json":"https://pith.science/pith/DBG5FU5Z7IA6ELRFXCGUDKM4PA.json","view_paper":"https://pith.science/paper/DBG5FU5Z","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2406.17957&json=true","fetch_graph":"https://pith.science/api/pith-number/DBG5FU5Z7IA6ELRFXCGUDKM4PA/graph.json","fetch_events":"https://pith.science/api/pith-number/DBG5FU5Z7IA6ELRFXCGUDKM4PA/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/DBG5FU5Z7IA6ELRFXCGUDKM4PA/action/timestamp_anchor","attest_storage":"https://pith.science/pith/DBG5FU5Z7IA6ELRFXCGUDKM4PA/action/storage_attestation","attest_author":"https://pith.science/pith/DBG5FU5Z7IA6ELRFXCGUDKM4PA/action/author_attestation","sign_citation":"https://pith.science/pith/DBG5FU5Z7IA6ELRFXCGUDKM4PA/action/citation_signature","submit_replication":"https://pith.science/pith/DBG5FU5Z7IA6ELRFXCGUDKM4PA/action/replication_record"}},"created_at":"2026-07-05T08:36:52.928708+00:00","updated_at":"2026-07-05T08:36:52.928708+00:00"}