{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:QOWGFSZFG6ZI37JCNZAUBRMDE2","merge_version":"pith-open-graph-merge-v1","event_count":2,"valid_event_count":2,"invalid_event_count":0,"equivocation_count":0,"current":{"canonical_record":{"metadata":{"abstract_canon_sha256":"09722db8d1bdf60f8e279b2cff0ff0db6ef4b09b01f7312f59cb279ad194359b","cross_cats_sorted":["eess.AS"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2026-05-18T00:22:18Z","title_canon_sha256":"93f4ca82ca769461639ef85d8e0280fe76d22b9b20b0df5e971c6b346f34650d"},"schema_version":"1.0","source":{"id":"2605.17710","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2605.17710","created_at":"2026-05-20T00:04:54Z"},{"alias_kind":"arxiv_version","alias_value":"2605.17710v1","created_at":"2026-05-20T00:04:54Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.17710","created_at":"2026-05-20T00:04:54Z"},{"alias_kind":"pith_short_12","alias_value":"QOWGFSZFG6ZI","created_at":"2026-05-20T00:04:54Z"},{"alias_kind":"pith_short_16","alias_value":"QOWGFSZFG6ZI37JC","created_at":"2026-05-20T00:04:54Z"},{"alias_kind":"pith_short_8","alias_value":"QOWGFSZF","created_at":"2026-05-20T00:04:54Z"}],"graph_snapshots":[{"event_id":"sha256:2b8310a98a5e89a20c5a063add4fc1d136733822a87e27a5370d435befde0b0b","target":"graph","created_at":"2026-05-20T00:04:54Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"graph_snapshot":{"author_claims":{"count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","strong_count":0},"builder_version":"pith-number-builder-2026-05-17-v1","claims":{"count":4,"items":[{"attestation":"unclaimed","claim_id":"C1","kind":"strongest_claim","source":"verdict.strongest_claim","status":"machine_extracted","text":"Our method significantly bridges the performance gap, achieving on average a relative Word Error Rate (WER) reduction of 29 % over monolingual baselines. Our models also outperform state-of-the-art multilingual models across major benchmarks, including Common Voice and Fleurs."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"The iterative self-improvement stage assumes that pseudo-labels generated by the initial distilled model are sufficiently accurate and do not introduce systematic errors that compound over iterations; this premise is invoked when describing the second stage of refinement on pseudo-labelled data."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"SBPN is a multilingual ASR model for five Nigerian languages that delivers a 29% relative WER reduction over monolingual baselines through teacher-student distillation plus iterative pseudo-label self-improvement."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"Knowledge distillation from monolingual models followed by self-improvement on pseudo labels improves multilingual ASR for Nigerian languages."}],"snapshot_sha256":"4bb3121599077e5c3e33b203c5d633fb28a0e4698027702ec04a6fda0fc4ddb4"},"formal_canon":{"evidence_count":2,"snapshot_sha256":"6be5b6b96be6767e2f5cdb9074d82164286f45b323d3d384c4c1a2982fdae5a2"},"integrity":{"available":true,"clean":true,"detectors_run":[{"findings_count":0,"name":"doi_title_agreement","ran_at":"2026-05-19T22:31:19.399565Z","status":"completed","version":"1.0.0"},{"findings_count":0,"name":"doi_compliance","ran_at":"2026-05-19T22:10:57.167441Z","status":"completed","version":"1.0.0"},{"findings_count":0,"name":"shingle_duplication","ran_at":"2026-05-19T21:49:43.475401Z","status":"skipped","version":"0.1.0"},{"findings_count":0,"name":"citation_quote_validity","ran_at":"2026-05-19T21:49:43.304918Z","status":"skipped","version":"0.1.0"},{"findings_count":0,"name":"ai_meta_artifact","ran_at":"2026-05-19T21:33:23.508691Z","status":"skipped","version":"1.0.0"},{"findings_count":0,"name":"cited_work_retraction","ran_at":"2026-05-19T21:21:58.548102Z","status":"completed","version":"1.0.0"},{"findings_count":0,"name":"claim_evidence","ran_at":"2026-05-19T21:21:57.414497Z","status":"completed","version":"1.0.0"}],"endpoint":"/pith/2605.17710/integrity.json","findings":[],"snapshot_sha256":"fa437c79dae3535521258ba9aa833e772f41544d169e0f1c79fad2aec525035b","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Although modern multilingual Automatic Speech Recognition (ASR) systems support several Nigerian languages, their performance consistently lags behind high-resource languages like English and French. Nigerian languages present unique modelling hurdles, including acute data scarcity, inconsistent orthography, tonal diacritics, diverse accents, frequent code-switching, and localized named entities. To address these challenges, we developed a multilingual ASR framework utilizing a two-stage distillation process. First, we employ student-teacher knowledge distillation from existing monolingual mod","authors_text":"Sewade Ogun","cross_cats":["eess.AS"],"headline":"Knowledge distillation from monolingual models followed by self-improvement on pseudo labels improves multilingual ASR for Nigerian languages.","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2026-05-18T00:22:18Z","title":"Sometin Beta Pass Notin (SBPN): Improving Multilingual ASR for Nigerian Languages via Knowledge Distillation"},"references":{"count":31,"internal_anchors":0,"resolved_work":31,"sample":[{"cited_arxiv_id":"","doi":"10.21437/interspeech.2023-466","is_internal_anchor":false,"ref_index":1,"title":"doi: 10.21437/Interspeech.2023-466. Josh Meyer, David Adelani, Edresson Casanova, Alp ¨Oktem, Daniel Whitenack, Julian Weber, Salomon KABONGO KABENAMUALU, Elizabeth Salesky, Iroro Orife, Colin Leong, ","work_id":"6b2f3708-8152-4404-b26e-701631d8e8af","year":2023},{"cited_arxiv_id":"","doi":"10.18653/v1/2024.emnlp-main.983","is_internal_anchor":false,"ref_index":2,"title":"In: Proceedings of the 2024 Conference on Em- pirical Methods in Natural Language Processing, EMNLP 2024, Miami, FL, USA, November 12-16","work_id":"c9b74554-3677-4c0a-87dc-5bb9b1a68e21","year":2024},{"cited_arxiv_id":"","doi":"10.1109/taslp.2023.3306709","is_internal_anchor":false,"ref_index":3,"title":"20 Sometin Beta Pass Notin (SBPN) Appendix A","work_id":"f51842b8-f8dc-41d9-abff-b088c3f3b057","year":2023},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":4,"title":"beforebefore→bifor bifor","work_id":"a61ec188-80dc-48c1-8543-b5dee0700156","year":null},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":5,"title":"chewing gum→ chingum","work_id":"b63824d7-7689-4e82-86dd-fd2465877ba3","year":null}],"snapshot_sha256":"468025f93fb317d7e0e0daa205c683259373c808cca120c4c396812dfde92d2d"},"source":{"id":"2605.17710","kind":"arxiv","version":1},"verdict":{"created_at":"2026-05-19T22:01:06.027531Z","id":"0a6a8fa7-91fd-4cb6-b831-1550e48d0a82","model_set":{"reader":"grok-4.3"},"one_line_summary":"SBPN is a multilingual ASR model for five Nigerian languages that delivers a 29% relative WER reduction over monolingual baselines through teacher-student distillation plus iterative pseudo-label self-improvement.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"Knowledge distillation from monolingual models followed by self-improvement on pseudo labels improves multilingual ASR for Nigerian languages.","strongest_claim":"Our method significantly bridges the performance gap, achieving on average a relative Word Error Rate (WER) reduction of 29 % over monolingual baselines. Our models also outperform state-of-the-art multilingual models across major benchmarks, including Common Voice and Fleurs.","weakest_assumption":"The iterative self-improvement stage assumes that pseudo-labels generated by the initial distilled model are sufficiently accurate and do not introduce systematic errors that compound over iterations; this premise is invoked when describing the second stage of refinement on pseudo-labelled data."}},"verdict_id":"0a6a8fa7-91fd-4cb6-b831-1550e48d0a82"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:d83f6ca14a3e400b38c3111c8abc0cd049dfdab389e086eb2fc7ca5b58d67544","target":"record","created_at":"2026-05-20T00:04:54Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"attestation_state":"computed","canonical_record":{"metadata":{"abstract_canon_sha256":"09722db8d1bdf60f8e279b2cff0ff0db6ef4b09b01f7312f59cb279ad194359b","cross_cats_sorted":["eess.AS"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2026-05-18T00:22:18Z","title_canon_sha256":"93f4ca82ca769461639ef85d8e0280fe76d22b9b20b0df5e971c6b346f34650d"},"schema_version":"1.0","source":{"id":"2605.17710","kind":"arxiv","version":1}},"canonical_sha256":"83ac62cb2537b28dfd226e4140c58326bb3060503e6a6722d3407674b9ea0527","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"83ac62cb2537b28dfd226e4140c58326bb3060503e6a6722d3407674b9ea0527","first_computed_at":"2026-05-20T00:04:54.103176Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-05-20T00:04:54.103176Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"UWil38vL5kkaQW0/AQCUP4GgCcGADfN+6rOXVELUQAZcIF2hPQ8jpoOfnbVj9LapDWk8PAGOSgTa5MypQqmqBQ==","signature_status":"signed_v1","signed_at":"2026-05-20T00:04:54.103946Z","signed_message":"canonical_sha256_bytes"},"source_id":"2605.17710","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:d83f6ca14a3e400b38c3111c8abc0cd049dfdab389e086eb2fc7ca5b58d67544","sha256:2b8310a98a5e89a20c5a063add4fc1d136733822a87e27a5370d435befde0b0b"],"state_sha256":"9827dfb7b123515448ab54e63850181b5a515a7134e64a28c47dc88b65aee6f6"}