{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2021:27CWYZFNUB7WDARAWLF2OPK4TX","short_pith_number":"pith:27CWYZFN","schema_version":"1.0","canonical_sha256":"d7c56c64ada07f618220b2cba73d5c9ddb9c08e04c838518d7115e795b3244d4","source":{"kind":"arxiv","id":"2110.02095","version":1},"attestation_state":"computed","paper":{"title":"Exploring the Limits of Large Scale Pre-training","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.CV","stat.ML"],"primary_cat":"cs.LG","authors_text":"Behnam Neyshabur, Hanie Sedghi, Mostafa Dehghani, Samira Abnar","submitted_at":"2021-10-05T14:49:00Z","abstract_excerpt":"Recent developments in large-scale machine learning suggest that by scaling up data, model size and training time properly, one might observe that improvements in pre-training would transfer favorably to most downstream tasks. In this work, we systematically study this phenomena and establish that, as we increase the upstream accuracy, the performance of downstream tasks saturates. In particular, we investigate more than 4800 experiments on Vision Transformers, MLP-Mixers and ResNets with number of parameters ranging from ten million to ten billion, trained on the largest scale of available im"},"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":"2110.02095","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2021-10-05T14:49:00Z","cross_cats_sorted":["cs.AI","cs.CV","stat.ML"],"title_canon_sha256":"061af57981582c122b378c3750e80e2d80eb5e606852bda3cf38df2d7f82773b","abstract_canon_sha256":"346b86c337248d00470731853565e1be2c5d60d5acc1330b8a0a16046757ba91"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T03:20:09.464790Z","signature_b64":"aX/PtnVf1MhTwMe+x2XEFlTlAPgEgut7ePzJCsauNW4s/79Pes/1TmC/3pcreUIbttCwoFXuWgu88lyqZxr0BQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"d7c56c64ada07f618220b2cba73d5c9ddb9c08e04c838518d7115e795b3244d4","last_reissued_at":"2026-07-05T03:20:09.464278Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T03:20:09.464278Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Exploring the Limits of Large Scale Pre-training","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.CV","stat.ML"],"primary_cat":"cs.LG","authors_text":"Behnam Neyshabur, Hanie Sedghi, Mostafa Dehghani, Samira Abnar","submitted_at":"2021-10-05T14:49:00Z","abstract_excerpt":"Recent developments in large-scale machine learning suggest that by scaling up data, model size and training time properly, one might observe that improvements in pre-training would transfer favorably to most downstream tasks. In this work, we systematically study this phenomena and establish that, as we increase the upstream accuracy, the performance of downstream tasks saturates. In particular, we investigate more than 4800 experiments on Vision Transformers, MLP-Mixers and ResNets with number of parameters ranging from ten million to ten billion, trained on the largest scale of available im"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2110.02095","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/2110.02095/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":"2110.02095","created_at":"2026-07-05T03:20:09.464342+00:00"},{"alias_kind":"arxiv_version","alias_value":"2110.02095v1","created_at":"2026-07-05T03:20:09.464342+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2110.02095","created_at":"2026-07-05T03:20:09.464342+00:00"},{"alias_kind":"pith_short_12","alias_value":"27CWYZFNUB7W","created_at":"2026-07-05T03:20:09.464342+00:00"},{"alias_kind":"pith_short_16","alias_value":"27CWYZFNUB7WDARA","created_at":"2026-07-05T03:20:09.464342+00:00"},{"alias_kind":"pith_short_8","alias_value":"27CWYZFN","created_at":"2026-07-05T03:20:09.464342+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2508.09691","citing_title":"PaCo-FR: Patch-Pixel Aligned End-to-End Codebook Learning for Facial Representation Pre-training","ref_index":1,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/27CWYZFNUB7WDARAWLF2OPK4TX","json":"https://pith.science/pith/27CWYZFNUB7WDARAWLF2OPK4TX.json","graph_json":"https://pith.science/api/pith-number/27CWYZFNUB7WDARAWLF2OPK4TX/graph.json","events_json":"https://pith.science/api/pith-number/27CWYZFNUB7WDARAWLF2OPK4TX/events.json","paper":"https://pith.science/paper/27CWYZFN"},"agent_actions":{"view_html":"https://pith.science/pith/27CWYZFNUB7WDARAWLF2OPK4TX","download_json":"https://pith.science/pith/27CWYZFNUB7WDARAWLF2OPK4TX.json","view_paper":"https://pith.science/paper/27CWYZFN","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2110.02095&json=true","fetch_graph":"https://pith.science/api/pith-number/27CWYZFNUB7WDARAWLF2OPK4TX/graph.json","fetch_events":"https://pith.science/api/pith-number/27CWYZFNUB7WDARAWLF2OPK4TX/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/27CWYZFNUB7WDARAWLF2OPK4TX/action/timestamp_anchor","attest_storage":"https://pith.science/pith/27CWYZFNUB7WDARAWLF2OPK4TX/action/storage_attestation","attest_author":"https://pith.science/pith/27CWYZFNUB7WDARAWLF2OPK4TX/action/author_attestation","sign_citation":"https://pith.science/pith/27CWYZFNUB7WDARAWLF2OPK4TX/action/citation_signature","submit_replication":"https://pith.science/pith/27CWYZFNUB7WDARAWLF2OPK4TX/action/replication_record"}},"created_at":"2026-07-05T03:20:09.464342+00:00","updated_at":"2026-07-05T03:20:09.464342+00:00"}