{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2022:BPVWIFB4MBOOWE7VKUWAEW45LK","short_pith_number":"pith:BPVWIFB4","schema_version":"1.0","canonical_sha256":"0beb64143c605ceb13f5552c025b9d5a966a640002d330512cda798c202c4a5c","source":{"kind":"arxiv","id":"2203.07259","version":3},"attestation_state":"computed","paper":{"title":"The Optimal BERT Surgeon: Scalable and Accurate Second-Order Pruning for Large Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Benjamin Fineran, Dan Alistarh, Daniel Campos, Eldar Kurtic, Elias Frantar, Mark Kurtz, Michael Goin, Tuan Nguyen","submitted_at":"2022-03-14T16:40:31Z","abstract_excerpt":"Transformer-based language models have become a key building block for natural language processing. While these models are extremely accurate, they can be too large and computationally intensive to run on standard deployments. A variety of compression methods, including distillation, quantization, structured and unstructured pruning are known to decrease model size and increase inference speed, with low accuracy loss. In this context, this paper's contributions are two-fold. We perform an in-depth study of the accuracy-compression trade-off for unstructured weight pruning of BERT models. We in"},"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":"2203.07259","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2022-03-14T16:40:31Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"755e7cdc1098cbd1f493caabdb1f7c0a0442d7434ff295381ae02692ef2a2ff7","abstract_canon_sha256":"1737eb710dd74d76864695240920f25df0f27e9d4536b7d5ad66bd9e0017d67c"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T05:07:40.221887Z","signature_b64":"8KBGcgIqqY3Bo4ptftoVfTOMjEurl6k66KmP5XGewDpPGhCoOnDY51M7tTDytYMb3oFOsIp+MJjEBo5RHm8YCw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"0beb64143c605ceb13f5552c025b9d5a966a640002d330512cda798c202c4a5c","last_reissued_at":"2026-07-05T05:07:40.221411Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T05:07:40.221411Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"The Optimal BERT Surgeon: Scalable and Accurate Second-Order Pruning for Large Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Benjamin Fineran, Dan Alistarh, Daniel Campos, Eldar Kurtic, Elias Frantar, Mark Kurtz, Michael Goin, Tuan Nguyen","submitted_at":"2022-03-14T16:40:31Z","abstract_excerpt":"Transformer-based language models have become a key building block for natural language processing. While these models are extremely accurate, they can be too large and computationally intensive to run on standard deployments. A variety of compression methods, including distillation, quantization, structured and unstructured pruning are known to decrease model size and increase inference speed, with low accuracy loss. In this context, this paper's contributions are two-fold. We perform an in-depth study of the accuracy-compression trade-off for unstructured weight pruning of BERT models. We in"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2203.07259","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/2203.07259/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":"2203.07259","created_at":"2026-07-05T05:07:40.221469+00:00"},{"alias_kind":"arxiv_version","alias_value":"2203.07259v3","created_at":"2026-07-05T05:07:40.221469+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2203.07259","created_at":"2026-07-05T05:07:40.221469+00:00"},{"alias_kind":"pith_short_12","alias_value":"BPVWIFB4MBOO","created_at":"2026-07-05T05:07:40.221469+00:00"},{"alias_kind":"pith_short_16","alias_value":"BPVWIFB4MBOOWE7V","created_at":"2026-07-05T05:07:40.221469+00:00"},{"alias_kind":"pith_short_8","alias_value":"BPVWIFB4","created_at":"2026-07-05T05:07:40.221469+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2310.02277","citing_title":"Junk DNA Hypothesis: Pruning Small Pre-Trained Weights Irreversibly and Monotonically Impairs \"Difficult\" Downstream Tasks in LLMs","ref_index":27,"is_internal_anchor":false},{"citing_arxiv_id":"2401.15077","citing_title":"EAGLE: Speculative Sampling Requires Rethinking Feature Uncertainty","ref_index":61,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/BPVWIFB4MBOOWE7VKUWAEW45LK","json":"https://pith.science/pith/BPVWIFB4MBOOWE7VKUWAEW45LK.json","graph_json":"https://pith.science/api/pith-number/BPVWIFB4MBOOWE7VKUWAEW45LK/graph.json","events_json":"https://pith.science/api/pith-number/BPVWIFB4MBOOWE7VKUWAEW45LK/events.json","paper":"https://pith.science/paper/BPVWIFB4"},"agent_actions":{"view_html":"https://pith.science/pith/BPVWIFB4MBOOWE7VKUWAEW45LK","download_json":"https://pith.science/pith/BPVWIFB4MBOOWE7VKUWAEW45LK.json","view_paper":"https://pith.science/paper/BPVWIFB4","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2203.07259&json=true","fetch_graph":"https://pith.science/api/pith-number/BPVWIFB4MBOOWE7VKUWAEW45LK/graph.json","fetch_events":"https://pith.science/api/pith-number/BPVWIFB4MBOOWE7VKUWAEW45LK/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/BPVWIFB4MBOOWE7VKUWAEW45LK/action/timestamp_anchor","attest_storage":"https://pith.science/pith/BPVWIFB4MBOOWE7VKUWAEW45LK/action/storage_attestation","attest_author":"https://pith.science/pith/BPVWIFB4MBOOWE7VKUWAEW45LK/action/author_attestation","sign_citation":"https://pith.science/pith/BPVWIFB4MBOOWE7VKUWAEW45LK/action/citation_signature","submit_replication":"https://pith.science/pith/BPVWIFB4MBOOWE7VKUWAEW45LK/action/replication_record"}},"created_at":"2026-07-05T05:07:40.221469+00:00","updated_at":"2026-07-05T05:07:40.221469+00:00"}