{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:4FGGSO3R67DSPTOJC2MWSAAMRQ","short_pith_number":"pith:4FGGSO3R","schema_version":"1.0","canonical_sha256":"e14c693b71f7c727cdc9169969000c8c1b794d4104f11259df16a147567fa712","source":{"kind":"arxiv","id":"2506.15912","version":1},"attestation_state":"computed","paper":{"title":"Early Attentive Sparsification Accelerates Neural Speech Transcription","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":["cs.CL","cs.SD","eess.AS"],"primary_cat":"cs.LG","authors_text":"Hesham Mostafa, Sayeh Sharify, Tristan Webb, Wanzin Yazar, Xin Wang, Zifei Xu","submitted_at":"2025-06-18T23:16:02Z","abstract_excerpt":"Transformer-based neural speech processing has achieved state-of-the-art performance. Since speech audio signals are known to be highly compressible, here we seek to accelerate neural speech transcription by time-domain signal sparsification early in the neural encoding stage, taking advantage of the interpretability of the self-attention mechanism in transformer audio encoders. With the Whisper family of models, we perform a systematic architecture search over the joint space of sparsification stage (a certain encoder layer) and compression ratio (sparsity). We found that the best resulting s"},"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":"2506.15912","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","primary_cat":"cs.LG","submitted_at":"2025-06-18T23:16:02Z","cross_cats_sorted":["cs.CL","cs.SD","eess.AS"],"title_canon_sha256":"ee18b28ca8587a1423fadd497b36858f3e5cb676fef327d21f54e228eb7d5fa7","abstract_canon_sha256":"d196b851e98d25ffd48bec4336577dfc8dd130e896d158d5bdf0fdeda102433e"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:24:26.370509Z","signature_b64":"7eQZ7rBBrM4lP2shvW62knpwtOqhgHs9R29/eXCYHRM+7UKiMqKLPTTGFSsHvpIMyyyFjIK5P0W9NOrAUB7FCQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"e14c693b71f7c727cdc9169969000c8c1b794d4104f11259df16a147567fa712","last_reissued_at":"2026-07-05T11:24:26.369961Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:24:26.369961Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Early Attentive Sparsification Accelerates Neural Speech Transcription","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":["cs.CL","cs.SD","eess.AS"],"primary_cat":"cs.LG","authors_text":"Hesham Mostafa, Sayeh Sharify, Tristan Webb, Wanzin Yazar, Xin Wang, Zifei Xu","submitted_at":"2025-06-18T23:16:02Z","abstract_excerpt":"Transformer-based neural speech processing has achieved state-of-the-art performance. Since speech audio signals are known to be highly compressible, here we seek to accelerate neural speech transcription by time-domain signal sparsification early in the neural encoding stage, taking advantage of the interpretability of the self-attention mechanism in transformer audio encoders. With the Whisper family of models, we perform a systematic architecture search over the joint space of sparsification stage (a certain encoder layer) and compression ratio (sparsity). We found that the best resulting s"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2506.15912","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/2506.15912/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":"2506.15912","created_at":"2026-07-05T11:24:26.370034+00:00"},{"alias_kind":"arxiv_version","alias_value":"2506.15912v1","created_at":"2026-07-05T11:24:26.370034+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2506.15912","created_at":"2026-07-05T11:24:26.370034+00:00"},{"alias_kind":"pith_short_12","alias_value":"4FGGSO3R67DS","created_at":"2026-07-05T11:24:26.370034+00:00"},{"alias_kind":"pith_short_16","alias_value":"4FGGSO3R67DSPTOJ","created_at":"2026-07-05T11:24:26.370034+00:00"},{"alias_kind":"pith_short_8","alias_value":"4FGGSO3R","created_at":"2026-07-05T11:24:26.370034+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.01483","citing_title":"MURMUR: An Efficient Inference System for Long-Form ASR","ref_index":32,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/4FGGSO3R67DSPTOJC2MWSAAMRQ","json":"https://pith.science/pith/4FGGSO3R67DSPTOJC2MWSAAMRQ.json","graph_json":"https://pith.science/api/pith-number/4FGGSO3R67DSPTOJC2MWSAAMRQ/graph.json","events_json":"https://pith.science/api/pith-number/4FGGSO3R67DSPTOJC2MWSAAMRQ/events.json","paper":"https://pith.science/paper/4FGGSO3R"},"agent_actions":{"view_html":"https://pith.science/pith/4FGGSO3R67DSPTOJC2MWSAAMRQ","download_json":"https://pith.science/pith/4FGGSO3R67DSPTOJC2MWSAAMRQ.json","view_paper":"https://pith.science/paper/4FGGSO3R","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2506.15912&json=true","fetch_graph":"https://pith.science/api/pith-number/4FGGSO3R67DSPTOJC2MWSAAMRQ/graph.json","fetch_events":"https://pith.science/api/pith-number/4FGGSO3R67DSPTOJC2MWSAAMRQ/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/4FGGSO3R67DSPTOJC2MWSAAMRQ/action/timestamp_anchor","attest_storage":"https://pith.science/pith/4FGGSO3R67DSPTOJC2MWSAAMRQ/action/storage_attestation","attest_author":"https://pith.science/pith/4FGGSO3R67DSPTOJC2MWSAAMRQ/action/author_attestation","sign_citation":"https://pith.science/pith/4FGGSO3R67DSPTOJC2MWSAAMRQ/action/citation_signature","submit_replication":"https://pith.science/pith/4FGGSO3R67DSPTOJC2MWSAAMRQ/action/replication_record"}},"created_at":"2026-07-05T11:24:26.370034+00:00","updated_at":"2026-07-05T11:24:26.370034+00:00"}