{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:J74AOHJ7GMPGSTQ56BDLSUPWDL","short_pith_number":"pith:J74AOHJ7","schema_version":"1.0","canonical_sha256":"4ff8071d3f331e694e1df046b951f61ad95f2f0aa4b3abee246f6eb4cd081b94","source":{"kind":"arxiv","id":"2407.03645","version":3},"attestation_state":"computed","paper":{"title":"Continual Learning Optimizations for Auto-regressive Decoder of Multilingual ASR systems","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.SD","eess.AS"],"primary_cat":"cs.CL","authors_text":"Chin Yuen Kwok, Eng Siong Chng, Jia Qi Yip","submitted_at":"2024-07-04T05:35:47Z","abstract_excerpt":"Continual Learning (CL) involves fine-tuning pre-trained models with new data while maintaining the performance on the pre-trained data. This is particularly relevant for expanding multilingual ASR (MASR) capabilities. However, existing CL methods, mainly designed for computer vision and reinforcement learning tasks, often yield sub-optimal results when directly applied to MASR. We hypothesise that this is because CL of the auto-regressive decoder in the MASR model is difficult. To verify this, we propose four optimizations on the decoder. They include decoder-layer gradient surgery, freezing "},"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":"2407.03645","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2024-07-04T05:35:47Z","cross_cats_sorted":["cs.SD","eess.AS"],"title_canon_sha256":"af660e7b9b8153ffa1f844cd92c01b9183b5d98480f2f92276135c595bf18110","abstract_canon_sha256":"54677e6dbf6722ea750eb020a60fe81185e8caa39e52d71088c91ec6a18cca27"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:12:29.549286Z","signature_b64":"CSAwgjsv/q7iw2AZo4L+GHx7FIL8CVK07QjR+NtZOW/fOEGCtD/laUHy9dC4qsr8D/TirE592Xl47snYgwMlBA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"4ff8071d3f331e694e1df046b951f61ad95f2f0aa4b3abee246f6eb4cd081b94","last_reissued_at":"2026-07-05T09:12:29.548764Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:12:29.548764Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Continual Learning Optimizations for Auto-regressive Decoder of Multilingual ASR systems","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.SD","eess.AS"],"primary_cat":"cs.CL","authors_text":"Chin Yuen Kwok, Eng Siong Chng, Jia Qi Yip","submitted_at":"2024-07-04T05:35:47Z","abstract_excerpt":"Continual Learning (CL) involves fine-tuning pre-trained models with new data while maintaining the performance on the pre-trained data. This is particularly relevant for expanding multilingual ASR (MASR) capabilities. However, existing CL methods, mainly designed for computer vision and reinforcement learning tasks, often yield sub-optimal results when directly applied to MASR. We hypothesise that this is because CL of the auto-regressive decoder in the MASR model is difficult. To verify this, we propose four optimizations on the decoder. They include decoder-layer gradient surgery, freezing "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2407.03645","kind":"arxiv","version":3},"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/2407.03645/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":"2407.03645","created_at":"2026-07-05T09:12:29.548833+00:00"},{"alias_kind":"arxiv_version","alias_value":"2407.03645v3","created_at":"2026-07-05T09:12:29.548833+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2407.03645","created_at":"2026-07-05T09:12:29.548833+00:00"},{"alias_kind":"pith_short_12","alias_value":"J74AOHJ7GMPG","created_at":"2026-07-05T09:12:29.548833+00:00"},{"alias_kind":"pith_short_16","alias_value":"J74AOHJ7GMPGSTQ5","created_at":"2026-07-05T09:12:29.548833+00:00"},{"alias_kind":"pith_short_8","alias_value":"J74AOHJ7","created_at":"2026-07-05T09:12:29.548833+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/J74AOHJ7GMPGSTQ56BDLSUPWDL","json":"https://pith.science/pith/J74AOHJ7GMPGSTQ56BDLSUPWDL.json","graph_json":"https://pith.science/api/pith-number/J74AOHJ7GMPGSTQ56BDLSUPWDL/graph.json","events_json":"https://pith.science/api/pith-number/J74AOHJ7GMPGSTQ56BDLSUPWDL/events.json","paper":"https://pith.science/paper/J74AOHJ7"},"agent_actions":{"view_html":"https://pith.science/pith/J74AOHJ7GMPGSTQ56BDLSUPWDL","download_json":"https://pith.science/pith/J74AOHJ7GMPGSTQ56BDLSUPWDL.json","view_paper":"https://pith.science/paper/J74AOHJ7","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2407.03645&json=true","fetch_graph":"https://pith.science/api/pith-number/J74AOHJ7GMPGSTQ56BDLSUPWDL/graph.json","fetch_events":"https://pith.science/api/pith-number/J74AOHJ7GMPGSTQ56BDLSUPWDL/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/J74AOHJ7GMPGSTQ56BDLSUPWDL/action/timestamp_anchor","attest_storage":"https://pith.science/pith/J74AOHJ7GMPGSTQ56BDLSUPWDL/action/storage_attestation","attest_author":"https://pith.science/pith/J74AOHJ7GMPGSTQ56BDLSUPWDL/action/author_attestation","sign_citation":"https://pith.science/pith/J74AOHJ7GMPGSTQ56BDLSUPWDL/action/citation_signature","submit_replication":"https://pith.science/pith/J74AOHJ7GMPGSTQ56BDLSUPWDL/action/replication_record"}},"created_at":"2026-07-05T09:12:29.548833+00:00","updated_at":"2026-07-05T09:12:29.548833+00:00"}