{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2023:Z36IXYVWXP7DYDAH7L3RVV4ASM","short_pith_number":"pith:Z36IXYVW","canonical_record":{"source":{"id":"2304.07193","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2023-04-14T15:12:19Z","cross_cats_sorted":[],"title_canon_sha256":"650bdb0e4b9cf04f35d5ffbb3fa19438e33c34f7bde30dd3ab5438509a38f7f9","abstract_canon_sha256":"fb853c5085b281d817539b4c9785236d6d0bb4607eaf6363024916d316e58be9"},"schema_version":"1.0"},"canonical_sha256":"cefc8be2b6bbfe3c0c07faf71ad78093161e0916eed8f5aa35ceda7d1dd0ede4","source":{"kind":"arxiv","id":"2304.07193","version":2},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2304.07193","created_at":"2026-07-05T07:40:21Z"},{"alias_kind":"arxiv_version","alias_value":"2304.07193v2","created_at":"2026-07-05T07:40:21Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2304.07193","created_at":"2026-07-05T07:40:21Z"},{"alias_kind":"pith_short_12","alias_value":"Z36IXYVWXP7D","created_at":"2026-07-05T07:40:21Z"},{"alias_kind":"pith_short_16","alias_value":"Z36IXYVWXP7DYDAH","created_at":"2026-07-05T07:40:21Z"},{"alias_kind":"pith_short_8","alias_value":"Z36IXYVW","created_at":"2026-07-05T07:40:21Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2023:Z36IXYVWXP7DYDAH7L3RVV4ASM","target":"record","payload":{"canonical_record":{"source":{"id":"2304.07193","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2023-04-14T15:12:19Z","cross_cats_sorted":[],"title_canon_sha256":"650bdb0e4b9cf04f35d5ffbb3fa19438e33c34f7bde30dd3ab5438509a38f7f9","abstract_canon_sha256":"fb853c5085b281d817539b4c9785236d6d0bb4607eaf6363024916d316e58be9"},"schema_version":"1.0"},"canonical_sha256":"cefc8be2b6bbfe3c0c07faf71ad78093161e0916eed8f5aa35ceda7d1dd0ede4","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T07:40:21.097430Z","signature_b64":"H9FQZyJeLdDV7JeSd8htvlR8/fcCU9O2Is23cyug9+0po5aLZeoNwdqPYioVaYkZRciQCvr0mx3EnSXOkRh1CA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"cefc8be2b6bbfe3c0c07faf71ad78093161e0916eed8f5aa35ceda7d1dd0ede4","last_reissued_at":"2026-07-05T07:40:21.096935Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T07:40:21.096935Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2304.07193","source_version":2,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T07:40:21Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"YKcNYDQDMSCpMEErIba9bco5XjK1gCOl7eiPsR46RJPcycChaQ2cJOeNBYcPKxCorXEor/rhzvF9jVT+S4GQBQ==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-04T23:19:40.469657Z"},"content_sha256":"26fe81a99739e68d5511dbc86473f9fd0887f6b4fddc404b87e26bdbd15e6783","schema_version":"1.0","event_id":"sha256:26fe81a99739e68d5511dbc86473f9fd0887f6b4fddc404b87e26bdbd15e6783"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2023:Z36IXYVWXP7DYDAH7L3RVV4ASM","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"DINOv2: Learning Robust Visual Features without Supervision","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"Curated self-supervised pretraining on 142M images yields frozen visual features that match or beat OpenCLIP across image and pixel tasks without finetuning.","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Alaaeldin El-Nouby, Armand Joulin, Daniel Haziza, Francisco Massa, Gabriel Synnaeve, Herv\\'e Jegou, Hu Xu, Huy Vo, Ishan Misra, Julien Mairal, Mahmoud Assran, Marc Szafraniec, Maxime Oquab, Michael Rabbat, Nicolas Ballas, Patrick Labatut, Pierre Fernandez, Piotr Bojanowski, Po-Yao Huang, Russell Howes, Shang-Wen Li, Th\\'eo Moutakanni, Timoth\\'ee Darcet, Vasil Khalidov, Vasu Sharma, Wojciech Galuba","submitted_at":"2023-04-14T15:12:19Z","abstract_excerpt":"The recent breakthroughs in natural language processing for model pretraining on large quantities of data have opened the way for similar foundation models in computer vision. These models could greatly simplify the use of images in any system by producing all-purpose visual features, i.e., features that work across image distributions and tasks without finetuning. This work shows that existing pretraining methods, especially self-supervised methods, can produce such features if trained on enough curated data from diverse sources. We revisit existing approaches and combine different techniques"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"Central claim unavailable from malformed reader payload.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"Load-bearing assumption unavailable from malformed reader payload.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"Pith review generated a malformed one-line summary.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Curated self-supervised pretraining on 142M images yields frozen visual features that match or beat OpenCLIP across image and pixel tasks without finetuning.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"2d4d6959d6c73b2d38bd9ca56129daa1ecf3ddc61150af2aadc37382fd01f5fb"},"source":{"id":"2304.07193","kind":"arxiv","version":2},"verdict":{"id":"dbe41d26-f1f4-449f-9cff-4e89d3d99484","model_set":{"reader":"claude-opus-4-7"},"created_at":"2026-05-09T01:50:51.976975Z","strongest_claim":"Central claim unavailable from malformed reader payload.","one_line_summary":"Pith review generated a malformed one-line summary.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"Load-bearing assumption unavailable from malformed reader payload.","pith_extraction_headline":"Curated self-supervised pretraining on 142M images yields frozen visual features that match or beat OpenCLIP across image and pixel tasks without finetuning."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2304.07193/integrity.json","findings":[],"available":true,"detectors_run":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938"},"references":{"count":37,"sample":[{"doi":"","year":2024,"title":"3For context, a full Boeing 777 return flight between London and New York corresponds to approximately 560 tCO2eq. 21 Published in Transactions on Machine Learning Research (01/2024) Mahmoud Assran, Q","work_id":"b3d89c91-9b9c-4fdf-b194-a2922422ee6f","ref_index":1,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":1902,"title":"MultiGrain: a unified image embedding for classes and instances","work_id":"54472c12-9ee6-4a11-86b6-d57d3cfa0459","ref_index":2,"cited_arxiv_id":"1902.05509","is_internal_anchor":false},{"doi":"","year":2006,"title":"Are we done with imagenet?arXiv preprint arXiv:2006.07159","work_id":"9efae043-283b-44ae-8324-207d3747f93f","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":null,"title":"On the Opportunities and Risks of Foundation Models","work_id":"a18039e9-928d-47c9-a836-32656a71bf71","ref_index":4,"cited_arxiv_id":"2108.07258","is_internal_anchor":true},{"doi":"","year":null,"title":"Symbolic discovery of optimization algorithms","work_id":"2151a7b4-dbf0-490e-8582-83731a6bc17c","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":37,"snapshot_sha256":"569d8ce749e6992ad2394908d6547345bfcf25705c9b3db0c1867ced9a39d373","internal_anchors":8},"formal_canon":{"evidence_count":3,"snapshot_sha256":"1797be2a1486df3e3cd7083580815c40a5a2b4dfaa5408eb605685f6eba4307c"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"verdict_id":"dbe41d26-f1f4-449f-9cff-4e89d3d99484"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T07:40:21Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"yqtvSqP3yrHi9fHRtplX4+tD10usGLNCEYDkuR+vuzB70IK0LhHAevy/n4Dojy/Y27ho5+sopx4UXoPoFdV3Cg==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-04T23:19:40.470570Z"},"content_sha256":"709bd6c1aa0a5e0aef68f8145328f6212b496fb4326c1e15b8d07ff5d895eb70","schema_version":"1.0","event_id":"sha256:709bd6c1aa0a5e0aef68f8145328f6212b496fb4326c1e15b8d07ff5d895eb70"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/Z36IXYVWXP7DYDAH7L3RVV4ASM/bundle.json","state_url":"https://pith.science/pith/Z36IXYVWXP7DYDAH7L3RVV4ASM/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/Z36IXYVWXP7DYDAH7L3RVV4ASM/bundle.json","status":"primary"}],"public_keys":[{"key_id":"pith-v1-2026-05","algorithm":"ed25519","format":"raw","public_key_b64":"stVStoiQhXFxp4s2pdzPNoqVNBMojDU/fJ2db5S3CbM=","public_key_hex":"b2d552b68890857171a78b36a5dccf368a953413288c353f7c9d9d6f94b709b3","fingerprint_sha256_b32_first128bits":"RVFV5Z2OI2J3ZUO7ERDEBCYNKS","fingerprint_sha256_hex":"8d4b5ee74e4693bcd1df2446408b0d54","rotates_at":null,"url":"https://pith.science/pith-signing-key.json","notes":"Pith uses this Ed25519 key to sign canonical record SHA-256 digests. Verify with: ed25519_verify(public_key, message=canonical_sha256_bytes, signature=base64decode(signature_b64))."}],"merge_version":"pith-open-graph-merge-v1","built_at":"2026-08-04T23:19:40Z","links":{"resolver":"https://pith.science/pith/Z36IXYVWXP7DYDAH7L3RVV4ASM","bundle":"https://pith.science/pith/Z36IXYVWXP7DYDAH7L3RVV4ASM/bundle.json","state":"https://pith.science/pith/Z36IXYVWXP7DYDAH7L3RVV4ASM/state.json","well_known_bundle":"https://pith.science/.well-known/pith/Z36IXYVWXP7DYDAH7L3RVV4ASM/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2023:Z36IXYVWXP7DYDAH7L3RVV4ASM","merge_version":"pith-open-graph-merge-v1","event_count":2,"valid_event_count":2,"invalid_event_count":0,"equivocation_count":0,"current":{"canonical_record":{"metadata":{"abstract_canon_sha256":"fb853c5085b281d817539b4c9785236d6d0bb4607eaf6363024916d316e58be9","cross_cats_sorted":[],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2023-04-14T15:12:19Z","title_canon_sha256":"650bdb0e4b9cf04f35d5ffbb3fa19438e33c34f7bde30dd3ab5438509a38f7f9"},"schema_version":"1.0","source":{"id":"2304.07193","kind":"arxiv","version":2}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2304.07193","created_at":"2026-07-05T07:40:21Z"},{"alias_kind":"arxiv_version","alias_value":"2304.07193v2","created_at":"2026-07-05T07:40:21Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2304.07193","created_at":"2026-07-05T07:40:21Z"},{"alias_kind":"pith_short_12","alias_value":"Z36IXYVWXP7D","created_at":"2026-07-05T07:40:21Z"},{"alias_kind":"pith_short_16","alias_value":"Z36IXYVWXP7DYDAH","created_at":"2026-07-05T07:40:21Z"},{"alias_kind":"pith_short_8","alias_value":"Z36IXYVW","created_at":"2026-07-05T07:40:21Z"}],"graph_snapshots":[{"event_id":"sha256:709bd6c1aa0a5e0aef68f8145328f6212b496fb4326c1e15b8d07ff5d895eb70","target":"graph","created_at":"2026-07-05T07:40:21Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"graph_snapshot":{"author_claims":{"count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","strong_count":0},"builder_version":"pith-number-builder-2026-05-17-v1","claims":{"count":4,"items":[{"attestation":"unclaimed","claim_id":"C1","kind":"strongest_claim","source":"verdict.strongest_claim","status":"machine_extracted","text":"Central claim unavailable from malformed reader payload."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"Load-bearing assumption unavailable from malformed reader payload."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"Pith review generated a malformed one-line summary."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"Curated self-supervised pretraining on 142M images yields frozen visual features that match or beat OpenCLIP across image and pixel tasks without finetuning."}],"snapshot_sha256":"2d4d6959d6c73b2d38bd9ca56129daa1ecf3ddc61150af2aadc37382fd01f5fb"},"formal_canon":{"evidence_count":3,"snapshot_sha256":"1797be2a1486df3e3cd7083580815c40a5a2b4dfaa5408eb605685f6eba4307c"},"integrity":{"available":true,"clean":true,"detectors_run":[],"endpoint":"/pith/2304.07193/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"The recent breakthroughs in natural language processing for model pretraining on large quantities of data have opened the way for similar foundation models in computer vision. These models could greatly simplify the use of images in any system by producing all-purpose visual features, i.e., features that work across image distributions and tasks without finetuning. This work shows that existing pretraining methods, especially self-supervised methods, can produce such features if trained on enough curated data from diverse sources. We revisit existing approaches and combine different techniques","authors_text":"Alaaeldin El-Nouby, Armand Joulin, Daniel Haziza, Francisco Massa, Gabriel Synnaeve, Herv\\'e Jegou, Hu Xu, Huy Vo, Ishan Misra, Julien Mairal, Mahmoud Assran, Marc Szafraniec, Maxime Oquab, Michael Rabbat, Nicolas Ballas, Patrick Labatut, Pierre Fernandez, Piotr Bojanowski, Po-Yao Huang, Russell Howes, Shang-Wen Li, Th\\'eo Moutakanni, Timoth\\'ee Darcet, Vasil Khalidov, Vasu Sharma, Wojciech Galuba","cross_cats":[],"headline":"Curated self-supervised pretraining on 142M images yields frozen visual features that match or beat OpenCLIP across image and pixel tasks without finetuning.","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2023-04-14T15:12:19Z","title":"DINOv2: Learning Robust Visual Features without Supervision"},"references":{"count":37,"internal_anchors":8,"resolved_work":37,"sample":[{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":1,"title":"3For context, a full Boeing 777 return flight between London and New York corresponds to approximately 560 tCO2eq. 21 Published in Transactions on Machine Learning Research (01/2024) Mahmoud Assran, Q","work_id":"b3d89c91-9b9c-4fdf-b194-a2922422ee6f","year":2024},{"cited_arxiv_id":"1902.05509","doi":"","is_internal_anchor":false,"ref_index":2,"title":"MultiGrain: a unified image embedding for classes and instances","work_id":"54472c12-9ee6-4a11-86b6-d57d3cfa0459","year":1902},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":3,"title":"Are we done with imagenet?arXiv preprint arXiv:2006.07159","work_id":"9efae043-283b-44ae-8324-207d3747f93f","year":2006},{"cited_arxiv_id":"2108.07258","doi":"","is_internal_anchor":true,"ref_index":4,"title":"On the Opportunities and Risks of Foundation Models","work_id":"a18039e9-928d-47c9-a836-32656a71bf71","year":null},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":5,"title":"Symbolic discovery of optimization algorithms","work_id":"2151a7b4-dbf0-490e-8582-83731a6bc17c","year":null}],"snapshot_sha256":"569d8ce749e6992ad2394908d6547345bfcf25705c9b3db0c1867ced9a39d373"},"source":{"id":"2304.07193","kind":"arxiv","version":2},"verdict":{"created_at":"2026-05-09T01:50:51.976975Z","id":"dbe41d26-f1f4-449f-9cff-4e89d3d99484","model_set":{"reader":"claude-opus-4-7"},"one_line_summary":"Pith review generated a malformed one-line summary.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"Curated self-supervised pretraining on 142M images yields frozen visual features that match or beat OpenCLIP across image and pixel tasks without finetuning.","strongest_claim":"Central claim unavailable from malformed reader payload.","weakest_assumption":"Load-bearing assumption unavailable from malformed reader payload."}},"verdict_id":"dbe41d26-f1f4-449f-9cff-4e89d3d99484"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:26fe81a99739e68d5511dbc86473f9fd0887f6b4fddc404b87e26bdbd15e6783","target":"record","created_at":"2026-07-05T07:40:21Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"attestation_state":"computed","canonical_record":{"metadata":{"abstract_canon_sha256":"fb853c5085b281d817539b4c9785236d6d0bb4607eaf6363024916d316e58be9","cross_cats_sorted":[],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2023-04-14T15:12:19Z","title_canon_sha256":"650bdb0e4b9cf04f35d5ffbb3fa19438e33c34f7bde30dd3ab5438509a38f7f9"},"schema_version":"1.0","source":{"id":"2304.07193","kind":"arxiv","version":2}},"canonical_sha256":"cefc8be2b6bbfe3c0c07faf71ad78093161e0916eed8f5aa35ceda7d1dd0ede4","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"cefc8be2b6bbfe3c0c07faf71ad78093161e0916eed8f5aa35ceda7d1dd0ede4","first_computed_at":"2026-07-05T07:40:21.096935Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T07:40:21.096935Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"H9FQZyJeLdDV7JeSd8htvlR8/fcCU9O2Is23cyug9+0po5aLZeoNwdqPYioVaYkZRciQCvr0mx3EnSXOkRh1CA==","signature_status":"signed_v1","signed_at":"2026-07-05T07:40:21.097430Z","signed_message":"canonical_sha256_bytes"},"source_id":"2304.07193","source_kind":"arxiv","source_version":2}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:26fe81a99739e68d5511dbc86473f9fd0887f6b4fddc404b87e26bdbd15e6783","sha256:709bd6c1aa0a5e0aef68f8145328f6212b496fb4326c1e15b8d07ff5d895eb70"],"state_sha256":"f14eb2aa6ef9a8567b94ce8bb07231090f2c333ff872b60ef199eb47cb71cdb6"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"63nyw/hpVydJykl8g86/IoUf+15a1baEG1WVcGIoSxR/k7yIyBnTTPyUr7gvvJqPATHl+WjgzIlxKNbQA6W3CQ==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-04T23:19:40.477971Z","bundle_sha256":"a2de16a3f0daf10e1b3c57159f570a55dc6ae01952016da56283b0cc94af11a8"}}