{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2018:IQU7IWWND2YUX773D7FODAJRWN","short_pith_number":"pith:IQU7IWWN","schema_version":"1.0","canonical_sha256":"4429f45acd1eb14bfffb1fcae18131b36aa452d0283e70a29c8f709a4c4c2a4d","source":{"kind":"arxiv","id":"1810.05270","version":2},"attestation_state":"computed","paper":{"title":"Rethinking the Value of Network Pruning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CV","stat.ML"],"primary_cat":"cs.LG","authors_text":"Gao Huang, Mingjie Sun, Tinghui Zhou, Trevor Darrell, Zhuang Liu","submitted_at":"2018-10-11T22:15:28Z","abstract_excerpt":"Network pruning is widely used for reducing the heavy inference cost of deep models in low-resource settings. A typical pruning algorithm is a three-stage pipeline, i.e., training (a large model), pruning and fine-tuning. During pruning, according to a certain criterion, redundant weights are pruned and important weights are kept to best preserve the accuracy. In this work, we make several surprising observations which contradict common beliefs. For all state-of-the-art structured pruning algorithms we examined, fine-tuning a pruned model only gives comparable or worse performance than trainin"},"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":"1810.05270","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2018-10-11T22:15:28Z","cross_cats_sorted":["cs.CV","stat.ML"],"title_canon_sha256":"90362052d493c3f27fdd63a86fa76e466af08e8bbbec9b0e5535d4a52df21fff","abstract_canon_sha256":"10996b5c8f4a54566fd28f9df321382d2bbf1dc54563d6943b5a73bdcca66dda"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-17T23:52:04.866681Z","signature_b64":"yETwOL9lYXccpEalxOonMGwjjnliSzfJu3c5uQ3oy4ywtih8DEES0tIL+48lTXqeCkS/1GY2/nln/X8ScGWLDw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"4429f45acd1eb14bfffb1fcae18131b36aa452d0283e70a29c8f709a4c4c2a4d","last_reissued_at":"2026-05-17T23:52:04.866055Z","signature_status":"signed_v1","first_computed_at":"2026-05-17T23:52:04.866055Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Rethinking the Value of Network Pruning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CV","stat.ML"],"primary_cat":"cs.LG","authors_text":"Gao Huang, Mingjie Sun, Tinghui Zhou, Trevor Darrell, Zhuang Liu","submitted_at":"2018-10-11T22:15:28Z","abstract_excerpt":"Network pruning is widely used for reducing the heavy inference cost of deep models in low-resource settings. A typical pruning algorithm is a three-stage pipeline, i.e., training (a large model), pruning and fine-tuning. During pruning, according to a certain criterion, redundant weights are pruned and important weights are kept to best preserve the accuracy. In this work, we make several surprising observations which contradict common beliefs. For all state-of-the-art structured pruning algorithms we examined, fine-tuning a pruned model only gives comparable or worse performance than trainin"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"1810.05270","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":""},"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":"1810.05270","created_at":"2026-05-17T23:52:04.866151+00:00"},{"alias_kind":"arxiv_version","alias_value":"1810.05270v2","created_at":"2026-05-17T23:52:04.866151+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.1810.05270","created_at":"2026-05-17T23:52:04.866151+00:00"},{"alias_kind":"pith_short_12","alias_value":"IQU7IWWND2YU","created_at":"2026-05-18T12:32:31.084164+00:00"},{"alias_kind":"pith_short_16","alias_value":"IQU7IWWND2YUX773","created_at":"2026-05-18T12:32:31.084164+00:00"},{"alias_kind":"pith_short_8","alias_value":"IQU7IWWN","created_at":"2026-05-18T12:32:31.084164+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":2,"sample":[{"citing_arxiv_id":"2306.14048","citing_title":"H$_2$O: Heavy-Hitter Oracle for Efficient Generative Inference of Large Language Models","ref_index":60,"is_internal_anchor":true},{"citing_arxiv_id":"2310.12508","citing_title":"SalUn: Empowering Machine Unlearning via Gradient-based Weight Saliency in Both Image Classification and Generation","ref_index":197,"is_internal_anchor":true},{"citing_arxiv_id":"2604.17476","citing_title":"Privatar: Scalable Privacy-preserving Multi-user VR via Secure Offloading","ref_index":115,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/IQU7IWWND2YUX773D7FODAJRWN","json":"https://pith.science/pith/IQU7IWWND2YUX773D7FODAJRWN.json","graph_json":"https://pith.science/api/pith-number/IQU7IWWND2YUX773D7FODAJRWN/graph.json","events_json":"https://pith.science/api/pith-number/IQU7IWWND2YUX773D7FODAJRWN/events.json","paper":"https://pith.science/paper/IQU7IWWN"},"agent_actions":{"view_html":"https://pith.science/pith/IQU7IWWND2YUX773D7FODAJRWN","download_json":"https://pith.science/pith/IQU7IWWND2YUX773D7FODAJRWN.json","view_paper":"https://pith.science/paper/IQU7IWWN","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=1810.05270&json=true","fetch_graph":"https://pith.science/api/pith-number/IQU7IWWND2YUX773D7FODAJRWN/graph.json","fetch_events":"https://pith.science/api/pith-number/IQU7IWWND2YUX773D7FODAJRWN/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/IQU7IWWND2YUX773D7FODAJRWN/action/timestamp_anchor","attest_storage":"https://pith.science/pith/IQU7IWWND2YUX773D7FODAJRWN/action/storage_attestation","attest_author":"https://pith.science/pith/IQU7IWWND2YUX773D7FODAJRWN/action/author_attestation","sign_citation":"https://pith.science/pith/IQU7IWWND2YUX773D7FODAJRWN/action/citation_signature","submit_replication":"https://pith.science/pith/IQU7IWWND2YUX773D7FODAJRWN/action/replication_record"}},"created_at":"2026-05-17T23:52:04.866151+00:00","updated_at":"2026-05-17T23:52:04.866151+00:00"}