{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:XSJMEROLVMXOIWPJAU5Y5CXM5P","short_pith_number":"pith:XSJMEROL","schema_version":"1.0","canonical_sha256":"bc92c245cbab2ee459e9053b8e8aecebf91e0bce1b6988458444cd129a61720f","source":{"kind":"arxiv","id":"2511.08093","version":2},"attestation_state":"computed","paper":{"title":"Quantizing Whisper-small: How design choices affect ASR performance","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL","cs.SD"],"primary_cat":"eess.AS","authors_text":"Andreas S{\\o}eborg Kirkedal, Arthur S\\\"ohler, Julian Irigoyen","submitted_at":"2025-11-11T10:47:00Z","abstract_excerpt":"Large speech recognition models like Whisper-small achieve high accuracy but are difficult to deploy on edge devices due to their high computational demand. To this end, we present a unified, cross-library evaluation of post-training quantization (PTQ) on Whisper-small that disentangles the impact of quantization scheme, method, granularity, and bit-width. Our study is based on four libraries: PyTorch, Optimum-Quanto, HQQ, and bitsandbytes. Experiments on LibriSpeech test-clean and test-other show that dynamic int8 quantization with Quanto offers the best trade-off, reducing model size by 57% "},"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":"2511.08093","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"eess.AS","submitted_at":"2025-11-11T10:47:00Z","cross_cats_sorted":["cs.CL","cs.SD"],"title_canon_sha256":"e290a9bb9be2a6243d088e993aa004a061eba0ddd40387ad755c0041a81106f3","abstract_canon_sha256":"21c6eb35345fb32d961fd5f06d9592733c85d1a238f69d29849dab5410f159c5"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-22T01:03:16.219812Z","signature_b64":"DwJ9hMzocvjJV7OnUI1ehbbBRKoZWI6hYfqRV3wOQNDaA10gRT/ve2cj4A3E3wjy//2s6yQLN+W7KdNxDlVIAw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"bc92c245cbab2ee459e9053b8e8aecebf91e0bce1b6988458444cd129a61720f","last_reissued_at":"2026-05-22T01:03:16.218816Z","signature_status":"signed_v1","first_computed_at":"2026-05-22T01:03:16.218816Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Quantizing Whisper-small: How design choices affect ASR performance","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL","cs.SD"],"primary_cat":"eess.AS","authors_text":"Andreas S{\\o}eborg Kirkedal, Arthur S\\\"ohler, Julian Irigoyen","submitted_at":"2025-11-11T10:47:00Z","abstract_excerpt":"Large speech recognition models like Whisper-small achieve high accuracy but are difficult to deploy on edge devices due to their high computational demand. To this end, we present a unified, cross-library evaluation of post-training quantization (PTQ) on Whisper-small that disentangles the impact of quantization scheme, method, granularity, and bit-width. Our study is based on four libraries: PyTorch, Optimum-Quanto, HQQ, and bitsandbytes. Experiments on LibriSpeech test-clean and test-other show that dynamic int8 quantization with Quanto offers the best trade-off, reducing model size by 57% "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2511.08093","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/2511.08093/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":"2511.08093","created_at":"2026-05-22T01:03:16.218988+00:00"},{"alias_kind":"arxiv_version","alias_value":"2511.08093v2","created_at":"2026-05-22T01:03:16.218988+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2511.08093","created_at":"2026-05-22T01:03:16.218988+00:00"},{"alias_kind":"pith_short_12","alias_value":"XSJMEROLVMXO","created_at":"2026-05-22T01:03:16.218988+00:00"},{"alias_kind":"pith_short_16","alias_value":"XSJMEROLVMXOIWPJ","created_at":"2026-05-22T01:03:16.218988+00:00"},{"alias_kind":"pith_short_8","alias_value":"XSJMEROL","created_at":"2026-05-22T01:03:16.218988+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2511.08093","citing_title":"Quantizing Whisper-small: How design choices affect ASR performance","ref_index":2,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/XSJMEROLVMXOIWPJAU5Y5CXM5P","json":"https://pith.science/pith/XSJMEROLVMXOIWPJAU5Y5CXM5P.json","graph_json":"https://pith.science/api/pith-number/XSJMEROLVMXOIWPJAU5Y5CXM5P/graph.json","events_json":"https://pith.science/api/pith-number/XSJMEROLVMXOIWPJAU5Y5CXM5P/events.json","paper":"https://pith.science/paper/XSJMEROL"},"agent_actions":{"view_html":"https://pith.science/pith/XSJMEROLVMXOIWPJAU5Y5CXM5P","download_json":"https://pith.science/pith/XSJMEROLVMXOIWPJAU5Y5CXM5P.json","view_paper":"https://pith.science/paper/XSJMEROL","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2511.08093&json=true","fetch_graph":"https://pith.science/api/pith-number/XSJMEROLVMXOIWPJAU5Y5CXM5P/graph.json","fetch_events":"https://pith.science/api/pith-number/XSJMEROLVMXOIWPJAU5Y5CXM5P/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/XSJMEROLVMXOIWPJAU5Y5CXM5P/action/timestamp_anchor","attest_storage":"https://pith.science/pith/XSJMEROLVMXOIWPJAU5Y5CXM5P/action/storage_attestation","attest_author":"https://pith.science/pith/XSJMEROLVMXOIWPJAU5Y5CXM5P/action/author_attestation","sign_citation":"https://pith.science/pith/XSJMEROLVMXOIWPJAU5Y5CXM5P/action/citation_signature","submit_replication":"https://pith.science/pith/XSJMEROLVMXOIWPJAU5Y5CXM5P/action/replication_record"}},"created_at":"2026-05-22T01:03:16.218988+00:00","updated_at":"2026-05-22T01:03:16.218988+00:00"}