{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:CVBEYW2U4TKXTAL3BRMQGMS3PA","short_pith_number":"pith:CVBEYW2U","schema_version":"1.0","canonical_sha256":"15424c5b54e4d579817b0c5903325b7812382bd1117903c45aafb9e536057aa7","source":{"kind":"arxiv","id":"2508.07048","version":2},"attestation_state":"computed","paper":{"title":"Whisfusion: Parallel ASR Decoding with Masked Diffusion","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.LG","eess.AS"],"primary_cat":"cs.SD","authors_text":"Heeju Jwa, Hyuk-jae Lee, Hyungon Ryu, Jongchan Kim, Junhyuk Ahn, Nam-Joon Kim, Siwon Park, Taegeun Yun, Taeyoun Kwon, Yoonchae Choi","submitted_at":"2025-08-09T17:20:54Z","abstract_excerpt":"Autoregressive (AR) encoder-decoder models dominate high-quality multilingual ASR, but their left-to-right decoders make inference latency scale with transcript length. A natural alternative, CTC-style non-autoregressive (NAR) systems avoid this bottleneck but their conditional independence assumption sacrifices transcript-level generative modeling. Masked diffusion language models (e.g., LLaDA, MDLM) offer a competitive NAR text-generation approach. We ask whether such models can bring NAR ASR into the accuracy regime of strong AR ASR systems while removing the left-to-right bottleneck. We pr"},"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":"2508.07048","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.SD","submitted_at":"2025-08-09T17:20:54Z","cross_cats_sorted":["cs.AI","cs.LG","eess.AS"],"title_canon_sha256":"e9f379c740165d34707b996c94f28a4202933f5fc09182244e3ba2c209328a61","abstract_canon_sha256":"acdbad1d86bb04f93f9fcd982ea457c19ae46efb28a61fbed4969d085e1dd223"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-10T01:09:20.852713Z","signature_b64":"3vTKt+L1GB03YaGl5nHG0ne7u23Yqvwo3ix+pmtgMmLpiaVG4Jw1ZjQiZaclIlW4yGpvCLWelrxIgIZE73KbCg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"15424c5b54e4d579817b0c5903325b7812382bd1117903c45aafb9e536057aa7","last_reissued_at":"2026-06-10T01:09:20.851606Z","signature_status":"signed_v1","first_computed_at":"2026-06-10T01:09:20.851606Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Whisfusion: Parallel ASR Decoding with Masked Diffusion","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.LG","eess.AS"],"primary_cat":"cs.SD","authors_text":"Heeju Jwa, Hyuk-jae Lee, Hyungon Ryu, Jongchan Kim, Junhyuk Ahn, Nam-Joon Kim, Siwon Park, Taegeun Yun, Taeyoun Kwon, Yoonchae Choi","submitted_at":"2025-08-09T17:20:54Z","abstract_excerpt":"Autoregressive (AR) encoder-decoder models dominate high-quality multilingual ASR, but their left-to-right decoders make inference latency scale with transcript length. A natural alternative, CTC-style non-autoregressive (NAR) systems avoid this bottleneck but their conditional independence assumption sacrifices transcript-level generative modeling. Masked diffusion language models (e.g., LLaDA, MDLM) offer a competitive NAR text-generation approach. We ask whether such models can bring NAR ASR into the accuracy regime of strong AR ASR systems while removing the left-to-right bottleneck. We pr"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2508.07048","kind":"arxiv","version":2},"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/2508.07048/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":"2508.07048","created_at":"2026-06-10T01:09:20.851736+00:00"},{"alias_kind":"arxiv_version","alias_value":"2508.07048v2","created_at":"2026-06-10T01:09:20.851736+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2508.07048","created_at":"2026-06-10T01:09:20.851736+00:00"},{"alias_kind":"pith_short_12","alias_value":"CVBEYW2U4TKX","created_at":"2026-06-10T01:09:20.851736+00:00"},{"alias_kind":"pith_short_16","alias_value":"CVBEYW2U4TKXTAL3","created_at":"2026-06-10T01:09:20.851736+00:00"},{"alias_kind":"pith_short_8","alias_value":"CVBEYW2U","created_at":"2026-06-10T01:09:20.851736+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2604.14001","citing_title":"Diffusion Language Models for Speech Recognition","ref_index":20,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/CVBEYW2U4TKXTAL3BRMQGMS3PA","json":"https://pith.science/pith/CVBEYW2U4TKXTAL3BRMQGMS3PA.json","graph_json":"https://pith.science/api/pith-number/CVBEYW2U4TKXTAL3BRMQGMS3PA/graph.json","events_json":"https://pith.science/api/pith-number/CVBEYW2U4TKXTAL3BRMQGMS3PA/events.json","paper":"https://pith.science/paper/CVBEYW2U"},"agent_actions":{"view_html":"https://pith.science/pith/CVBEYW2U4TKXTAL3BRMQGMS3PA","download_json":"https://pith.science/pith/CVBEYW2U4TKXTAL3BRMQGMS3PA.json","view_paper":"https://pith.science/paper/CVBEYW2U","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2508.07048&json=true","fetch_graph":"https://pith.science/api/pith-number/CVBEYW2U4TKXTAL3BRMQGMS3PA/graph.json","fetch_events":"https://pith.science/api/pith-number/CVBEYW2U4TKXTAL3BRMQGMS3PA/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/CVBEYW2U4TKXTAL3BRMQGMS3PA/action/timestamp_anchor","attest_storage":"https://pith.science/pith/CVBEYW2U4TKXTAL3BRMQGMS3PA/action/storage_attestation","attest_author":"https://pith.science/pith/CVBEYW2U4TKXTAL3BRMQGMS3PA/action/author_attestation","sign_citation":"https://pith.science/pith/CVBEYW2U4TKXTAL3BRMQGMS3PA/action/citation_signature","submit_replication":"https://pith.science/pith/CVBEYW2U4TKXTAL3BRMQGMS3PA/action/replication_record"}},"created_at":"2026-06-10T01:09:20.851736+00:00","updated_at":"2026-06-10T01:09:20.851736+00:00"}