{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:7NUQYB4LRNAD6NNDCLYDIBXYZZ","short_pith_number":"pith:7NUQYB4L","schema_version":"1.0","canonical_sha256":"fb690c078b8b403f35a312f03406f8ce7af8353a7cd936159c97969256b6e68f","source":{"kind":"arxiv","id":"2302.01647","version":2},"attestation_state":"computed","paper":{"title":"Blockwise Self-Supervised Learning at Scale","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CV","authors_text":"David Krueger, Shoaib Ahmed Siddiqui, St\\'ephane Deny, Yann LeCun","submitted_at":"2023-02-03T10:48:24Z","abstract_excerpt":"Current state-of-the-art deep networks are all powered by backpropagation. In this paper, we explore alternatives to full backpropagation in the form of blockwise learning rules, leveraging the latest developments in self-supervised learning. We show that a blockwise pretraining procedure consisting of training independently the 4 main blocks of layers of a ResNet-50 with Barlow Twins' loss function at each block performs almost as well as end-to-end backpropagation on ImageNet: a linear probe trained on top of our blockwise pretrained model obtains a top-1 classification accuracy of 70.48%, o"},"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":"2302.01647","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","primary_cat":"cs.CV","submitted_at":"2023-02-03T10:48:24Z","cross_cats_sorted":["cs.AI","cs.LG"],"title_canon_sha256":"e8f1ba02e2b0bbd2f5930e7ca655a2ec1348e52a720200c048a50acc97c2aafc","abstract_canon_sha256":"f3e3dc5ad20ae94f057c8681310f48108363dee21f0e8c39c2dcb839d4018909"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:54:01.147693Z","signature_b64":"judEaZ8wFaN/YLZVVBJ7T9ZPo8r4A1NK9760Hb04VGLNlTwWArdQBAx28jgsdIiQV8b9hToGyx24zZxqK9raBA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"fb690c078b8b403f35a312f03406f8ce7af8353a7cd936159c97969256b6e68f","last_reissued_at":"2026-07-05T08:54:01.147260Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:54:01.147260Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Blockwise Self-Supervised Learning at Scale","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CV","authors_text":"David Krueger, Shoaib Ahmed Siddiqui, St\\'ephane Deny, Yann LeCun","submitted_at":"2023-02-03T10:48:24Z","abstract_excerpt":"Current state-of-the-art deep networks are all powered by backpropagation. In this paper, we explore alternatives to full backpropagation in the form of blockwise learning rules, leveraging the latest developments in self-supervised learning. We show that a blockwise pretraining procedure consisting of training independently the 4 main blocks of layers of a ResNet-50 with Barlow Twins' loss function at each block performs almost as well as end-to-end backpropagation on ImageNet: a linear probe trained on top of our blockwise pretrained model obtains a top-1 classification accuracy of 70.48%, o"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2302.01647","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/2302.01647/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":"2302.01647","created_at":"2026-07-05T08:54:01.147322+00:00"},{"alias_kind":"arxiv_version","alias_value":"2302.01647v2","created_at":"2026-07-05T08:54:01.147322+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2302.01647","created_at":"2026-07-05T08:54:01.147322+00:00"},{"alias_kind":"pith_short_12","alias_value":"7NUQYB4LRNAD","created_at":"2026-07-05T08:54:01.147322+00:00"},{"alias_kind":"pith_short_16","alias_value":"7NUQYB4LRNAD6NND","created_at":"2026-07-05T08:54:01.147322+00:00"},{"alias_kind":"pith_short_8","alias_value":"7NUQYB4L","created_at":"2026-07-05T08:54:01.147322+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2506.18193","citing_title":"DeInfoReg: A Decoupled Learning Framework for Better Training Throughput","ref_index":29,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/7NUQYB4LRNAD6NNDCLYDIBXYZZ","json":"https://pith.science/pith/7NUQYB4LRNAD6NNDCLYDIBXYZZ.json","graph_json":"https://pith.science/api/pith-number/7NUQYB4LRNAD6NNDCLYDIBXYZZ/graph.json","events_json":"https://pith.science/api/pith-number/7NUQYB4LRNAD6NNDCLYDIBXYZZ/events.json","paper":"https://pith.science/paper/7NUQYB4L"},"agent_actions":{"view_html":"https://pith.science/pith/7NUQYB4LRNAD6NNDCLYDIBXYZZ","download_json":"https://pith.science/pith/7NUQYB4LRNAD6NNDCLYDIBXYZZ.json","view_paper":"https://pith.science/paper/7NUQYB4L","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2302.01647&json=true","fetch_graph":"https://pith.science/api/pith-number/7NUQYB4LRNAD6NNDCLYDIBXYZZ/graph.json","fetch_events":"https://pith.science/api/pith-number/7NUQYB4LRNAD6NNDCLYDIBXYZZ/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/7NUQYB4LRNAD6NNDCLYDIBXYZZ/action/timestamp_anchor","attest_storage":"https://pith.science/pith/7NUQYB4LRNAD6NNDCLYDIBXYZZ/action/storage_attestation","attest_author":"https://pith.science/pith/7NUQYB4LRNAD6NNDCLYDIBXYZZ/action/author_attestation","sign_citation":"https://pith.science/pith/7NUQYB4LRNAD6NNDCLYDIBXYZZ/action/citation_signature","submit_replication":"https://pith.science/pith/7NUQYB4LRNAD6NNDCLYDIBXYZZ/action/replication_record"}},"created_at":"2026-07-05T08:54:01.147322+00:00","updated_at":"2026-07-05T08:54:01.147322+00:00"}