{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2025:5PDL7RXM5JEW3RYWA6AJRBHH3Z","short_pith_number":"pith:5PDL7RXM","canonical_record":{"source":{"id":"2508.10104","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2025-08-13T18:00:55Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"a4e398e2be4b48175c393b3fb3acd0236a9ab74855d9603f33e0fc9b1ae77ab9","abstract_canon_sha256":"0a435b2cfdaf55afc2cb20fe409d920c31a0ab4e68b02b2d38e96c6a823f7595"},"schema_version":"1.0"},"canonical_sha256":"ebc6bfc6ecea496dc71607809884e7de6847d561694e068e0bf8551baac1f126","source":{"kind":"arxiv","id":"2508.10104","version":1},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2508.10104","created_at":"2026-07-05T11:53:30Z"},{"alias_kind":"arxiv_version","alias_value":"2508.10104v1","created_at":"2026-07-05T11:53:30Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2508.10104","created_at":"2026-07-05T11:53:30Z"},{"alias_kind":"pith_short_12","alias_value":"5PDL7RXM5JEW","created_at":"2026-07-05T11:53:30Z"},{"alias_kind":"pith_short_16","alias_value":"5PDL7RXM5JEW3RYW","created_at":"2026-07-05T11:53:30Z"},{"alias_kind":"pith_short_8","alias_value":"5PDL7RXM","created_at":"2026-07-05T11:53:30Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2025:5PDL7RXM5JEW3RYWA6AJRBHH3Z","target":"record","payload":{"canonical_record":{"source":{"id":"2508.10104","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2025-08-13T18:00:55Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"a4e398e2be4b48175c393b3fb3acd0236a9ab74855d9603f33e0fc9b1ae77ab9","abstract_canon_sha256":"0a435b2cfdaf55afc2cb20fe409d920c31a0ab4e68b02b2d38e96c6a823f7595"},"schema_version":"1.0"},"canonical_sha256":"ebc6bfc6ecea496dc71607809884e7de6847d561694e068e0bf8551baac1f126","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:53:30.986833Z","signature_b64":"UO5DgQQ7ivZuE/yO/I2GnuH1vrxdLN7qRzCdZZppyUG3mVrghTCUzcT7jcofwamVTg6Ny6OfM+uBbsdKxTLnBg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"ebc6bfc6ecea496dc71607809884e7de6847d561694e068e0bf8551baac1f126","last_reissued_at":"2026-07-05T11:53:30.986327Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:53:30.986327Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2508.10104","source_version":1,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T11:53:30Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"tOB2gHo+4/tq3wfg+ZKoyPkVG6u+ErUgpiSY6b0VYzfHQXPN+Kj3W66axAW12KbVsdoo8oPk1D128L0epDdfDg==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-04T19:58:42.440859Z"},"content_sha256":"30abe0c4878a7bc927bbce1e013c7a952eacfb15ed35f993fe9827e5eb99b325","schema_version":"1.0","event_id":"sha256:30abe0c4878a7bc927bbce1e013c7a952eacfb15ed35f993fe9827e5eb99b325"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2025:5PDL7RXM5JEW3RYWA6AJRBHH3Z","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"DINOv3","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"DINOv3 uses Gram anchoring to keep dense visual features sharp through long self-supervised training.","cross_cats":["cs.LG"],"primary_cat":"cs.CV","authors_text":"Andrea Vedaldi, Camille Couprie, Cijo Jose, Claire Roberts, Daniel Haziza, Federico Baldassarre, Francisco Massa, Herv\\'e J\\'egou, Huy V. Vo, Jamie Tolan, Jianyuan Wang, John Brandt, Julien Mairal, Leonel Sentana, Luca Wehrstedt, Marc Szafraniec, Maxime Oquab, Maximilian Seitzer, Micha\\\"el Ramamonjisoa, Oriane Sim\\'eoni, Patrick Labatut, Piotr Bojanowski, Seungeun Yi, Th\\'eo Moutakanni, Timoth\\'ee Darcet, Vasil Khalidov","submitted_at":"2025-08-13T18:00:55Z","abstract_excerpt":"Self-supervised learning holds the promise of eliminating the need for manual data annotation, enabling models to scale effortlessly to massive datasets and larger architectures. By not being tailored to specific tasks or domains, this training paradigm has the potential to learn visual representations from diverse sources, ranging from natural to aerial images -- using a single algorithm. This technical report introduces DINOv3, a major milestone toward realizing this vision by leveraging simple yet effective strategies. First, we leverage the benefit of scaling both dataset and model size by"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"we present a versatile vision foundation model that outperforms the specialized state of the art across a broad range of settings, without fine-tuning. DINOv3 produces high-quality dense features that achieve outstanding performance on various vision tasks, significantly surpassing previous self- and weakly-supervised foundation models.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"The assumption that Gram anchoring successfully prevents degradation of dense feature maps over long training schedules, as this is presented as the key fix enabling the scaling benefits described in the abstract.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"DINOv3 advances self-supervised vision foundation models through dataset and model scaling, a new Gram anchoring technique for stable dense features, and post-hoc enhancements for resolution, size, and text alignment, claiming to outperform prior self- and weakly-supervised models without fine-tuned","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"DINOv3 uses Gram anchoring to keep dense visual features sharp through long self-supervised training.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"a189ccc3f558e75a7edcab0ecf57263c32e2a3922c749f6ea85ea20465a224c3"},"source":{"id":"2508.10104","kind":"arxiv","version":1},"verdict":{"id":"7dff7386-e4fc-4166-864c-6d4af6b20343","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-18T22:31:04.955065Z","strongest_claim":"we present a versatile vision foundation model that outperforms the specialized state of the art across a broad range of settings, without fine-tuning. DINOv3 produces high-quality dense features that achieve outstanding performance on various vision tasks, significantly surpassing previous self- and weakly-supervised foundation models.","one_line_summary":"DINOv3 advances self-supervised vision foundation models through dataset and model scaling, a new Gram anchoring technique for stable dense features, and post-hoc enhancements for resolution, size, and text alignment, claiming to outperform prior self- and weakly-supervised models without fine-tuned","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"The assumption that Gram anchoring successfully prevents degradation of dense feature maps over long training schedules, as this is presented as the key fix enabling the scaling benefits described in the abstract.","pith_extraction_headline":"DINOv3 uses Gram anchoring to keep dense visual features sharp through long self-supervised training."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2508.10104/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":2,"snapshot_sha256":"2e5f5a94c3614dfb1ad2a1854c73864bdf8c5eab7b4d7b4c68e3ffb7b571e550"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"verdict_id":"7dff7386-e4fc-4166-864c-6d4af6b20343"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T11:53:30Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"TNG2yKHOmL4oPXe2vg45N6zg1UVr/WDkPrWH8Dk9Bw6FK6DARpuDICRittYTxbXNn0NF5NWwJiAHVjg3rrODDQ==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-04T19:58:42.446160Z"},"content_sha256":"1e2501937c0082647f20aae2b4fd679155ec94745fb863dd982fcd75b80b49a8","schema_version":"1.0","event_id":"sha256:1e2501937c0082647f20aae2b4fd679155ec94745fb863dd982fcd75b80b49a8"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/5PDL7RXM5JEW3RYWA6AJRBHH3Z/bundle.json","state_url":"https://pith.science/pith/5PDL7RXM5JEW3RYWA6AJRBHH3Z/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/5PDL7RXM5JEW3RYWA6AJRBHH3Z/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":"2026-08-04T19:58:42Z","links":{"resolver":"https://pith.science/pith/5PDL7RXM5JEW3RYWA6AJRBHH3Z","bundle":"https://pith.science/pith/5PDL7RXM5JEW3RYWA6AJRBHH3Z/bundle.json","state":"https://pith.science/pith/5PDL7RXM5JEW3RYWA6AJRBHH3Z/state.json","well_known_bundle":"https://pith.science/.well-known/pith/5PDL7RXM5JEW3RYWA6AJRBHH3Z/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2025:5PDL7RXM5JEW3RYWA6AJRBHH3Z","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":"0a435b2cfdaf55afc2cb20fe409d920c31a0ab4e68b02b2d38e96c6a823f7595","cross_cats_sorted":["cs.LG"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2025-08-13T18:00:55Z","title_canon_sha256":"a4e398e2be4b48175c393b3fb3acd0236a9ab74855d9603f33e0fc9b1ae77ab9"},"schema_version":"1.0","source":{"id":"2508.10104","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2508.10104","created_at":"2026-07-05T11:53:30Z"},{"alias_kind":"arxiv_version","alias_value":"2508.10104v1","created_at":"2026-07-05T11:53:30Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2508.10104","created_at":"2026-07-05T11:53:30Z"},{"alias_kind":"pith_short_12","alias_value":"5PDL7RXM5JEW","created_at":"2026-07-05T11:53:30Z"},{"alias_kind":"pith_short_16","alias_value":"5PDL7RXM5JEW3RYW","created_at":"2026-07-05T11:53:30Z"},{"alias_kind":"pith_short_8","alias_value":"5PDL7RXM","created_at":"2026-07-05T11:53:30Z"}],"graph_snapshots":[{"event_id":"sha256:1e2501937c0082647f20aae2b4fd679155ec94745fb863dd982fcd75b80b49a8","target":"graph","created_at":"2026-07-05T11:53:30Z","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":"we present a versatile vision foundation model that outperforms the specialized state of the art across a broad range of settings, without fine-tuning. DINOv3 produces high-quality dense features that achieve outstanding performance on various vision tasks, significantly surpassing previous self- and weakly-supervised foundation models."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"The assumption that Gram anchoring successfully prevents degradation of dense feature maps over long training schedules, as this is presented as the key fix enabling the scaling benefits described in the abstract."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"DINOv3 advances self-supervised vision foundation models through dataset and model scaling, a new Gram anchoring technique for stable dense features, and post-hoc enhancements for resolution, size, and text alignment, claiming to outperform prior self- and weakly-supervised models without fine-tuned"},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"DINOv3 uses Gram anchoring to keep dense visual features sharp through long self-supervised training."}],"snapshot_sha256":"a189ccc3f558e75a7edcab0ecf57263c32e2a3922c749f6ea85ea20465a224c3"},"formal_canon":{"evidence_count":2,"snapshot_sha256":"2e5f5a94c3614dfb1ad2a1854c73864bdf8c5eab7b4d7b4c68e3ffb7b571e550"},"integrity":{"available":true,"clean":true,"detectors_run":[],"endpoint":"/pith/2508.10104/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Self-supervised learning holds the promise of eliminating the need for manual data annotation, enabling models to scale effortlessly to massive datasets and larger architectures. By not being tailored to specific tasks or domains, this training paradigm has the potential to learn visual representations from diverse sources, ranging from natural to aerial images -- using a single algorithm. This technical report introduces DINOv3, a major milestone toward realizing this vision by leveraging simple yet effective strategies. First, we leverage the benefit of scaling both dataset and model size by","authors_text":"Andrea Vedaldi, Camille Couprie, Cijo Jose, Claire Roberts, Daniel Haziza, Federico Baldassarre, Francisco Massa, Herv\\'e J\\'egou, Huy V. Vo, Jamie Tolan, Jianyuan Wang, John Brandt, Julien Mairal, Leonel Sentana, Luca Wehrstedt, Marc Szafraniec, Maxime Oquab, Maximilian Seitzer, Micha\\\"el Ramamonjisoa, Oriane Sim\\'eoni, Patrick Labatut, Piotr Bojanowski, Seungeun Yi, Th\\'eo Moutakanni, Timoth\\'ee Darcet, Vasil Khalidov","cross_cats":["cs.LG"],"headline":"DINOv3 uses Gram anchoring to keep dense visual features sharp through long self-supervised training.","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2025-08-13T18:00:55Z","title":"DINOv3"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2508.10104","kind":"arxiv","version":1},"verdict":{"created_at":"2026-05-18T22:31:04.955065Z","id":"7dff7386-e4fc-4166-864c-6d4af6b20343","model_set":{"reader":"grok-4.3"},"one_line_summary":"DINOv3 advances self-supervised vision foundation models through dataset and model scaling, a new Gram anchoring technique for stable dense features, and post-hoc enhancements for resolution, size, and text alignment, claiming to outperform prior self- and weakly-supervised models without fine-tuned","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"DINOv3 uses Gram anchoring to keep dense visual features sharp through long self-supervised training.","strongest_claim":"we present a versatile vision foundation model that outperforms the specialized state of the art across a broad range of settings, without fine-tuning. DINOv3 produces high-quality dense features that achieve outstanding performance on various vision tasks, significantly surpassing previous self- and weakly-supervised foundation models.","weakest_assumption":"The assumption that Gram anchoring successfully prevents degradation of dense feature maps over long training schedules, as this is presented as the key fix enabling the scaling benefits described in the abstract."}},"verdict_id":"7dff7386-e4fc-4166-864c-6d4af6b20343"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:30abe0c4878a7bc927bbce1e013c7a952eacfb15ed35f993fe9827e5eb99b325","target":"record","created_at":"2026-07-05T11:53:30Z","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":"0a435b2cfdaf55afc2cb20fe409d920c31a0ab4e68b02b2d38e96c6a823f7595","cross_cats_sorted":["cs.LG"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2025-08-13T18:00:55Z","title_canon_sha256":"a4e398e2be4b48175c393b3fb3acd0236a9ab74855d9603f33e0fc9b1ae77ab9"},"schema_version":"1.0","source":{"id":"2508.10104","kind":"arxiv","version":1}},"canonical_sha256":"ebc6bfc6ecea496dc71607809884e7de6847d561694e068e0bf8551baac1f126","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"ebc6bfc6ecea496dc71607809884e7de6847d561694e068e0bf8551baac1f126","first_computed_at":"2026-07-05T11:53:30.986327Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T11:53:30.986327Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"UO5DgQQ7ivZuE/yO/I2GnuH1vrxdLN7qRzCdZZppyUG3mVrghTCUzcT7jcofwamVTg6Ny6OfM+uBbsdKxTLnBg==","signature_status":"signed_v1","signed_at":"2026-07-05T11:53:30.986833Z","signed_message":"canonical_sha256_bytes"},"source_id":"2508.10104","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:30abe0c4878a7bc927bbce1e013c7a952eacfb15ed35f993fe9827e5eb99b325","sha256:1e2501937c0082647f20aae2b4fd679155ec94745fb863dd982fcd75b80b49a8"],"state_sha256":"fc3f06aba302653526fa7d0ccacb7078d5a6854ab4faf3e2c26d84469c838edb"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"BAsy7q6sFwXZumU63yaWc5r67Box63uGJlOl/221Bg2LiGGDfWDNUcFh/OZA6B22KkK8Dk2aVvi87m+fE71jAA==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-04T19:58:42.492021Z","bundle_sha256":"d1c88a3d3420f92f29f6ab8772e5cab9cb59ba373a2389e4a5ec1d2f0de47103"}}