{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:JY7C5FA46RMWK4V7PJDQH3WHC4","short_pith_number":"pith:JY7C5FA4","schema_version":"1.0","canonical_sha256":"4e3e2e941cf4596572bf7a4703eec7172ea46e83f45fe1cb9ab8414882852fda","source":{"kind":"arxiv","id":"2304.10520","version":2},"attestation_state":"computed","paper":{"title":"Contrastive Tuning: A Little Help to Make Masked Autoencoders Forget","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CV","authors_text":"Andreas F\\\"urst, Benedikt Alkin, Elisabeth Rumetshofer, Johannes Lehner, Lukas Miklautz, Sepp Hochreiter","submitted_at":"2023-04-20T17:51:09Z","abstract_excerpt":"Masked Image Modeling (MIM) methods, like Masked Autoencoders (MAE), efficiently learn a rich representation of the input. However, for adapting to downstream tasks, they require a sufficient amount of labeled data since their rich features code not only objects but also less relevant image background. In contrast, Instance Discrimination (ID) methods focus on objects. In this work, we study how to combine the efficiency and scalability of MIM with the ability of ID to perform downstream classification in the absence of large amounts of labeled data. To this end, we introduce Masked Autoencode"},"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":"2304.10520","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2023-04-20T17:51:09Z","cross_cats_sorted":["cs.AI","cs.LG"],"title_canon_sha256":"66a743a908039533f59134cf5cf60c3a1ed87c9506322a044c255de167f878a8","abstract_canon_sha256":"92aaa38cccef0738493febda8892ff47fe784f2ef1b53f04b4aceb86749a8912"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T06:50:40.202831Z","signature_b64":"aKJIoeEWRoKLGwvPfAa0+s9i14fDcJt4u5cMaXFGH+ji9xgsNejfb0oP7D4/zTp/QuGJYKXj13TreCg/rhz4Bw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"4e3e2e941cf4596572bf7a4703eec7172ea46e83f45fe1cb9ab8414882852fda","last_reissued_at":"2026-07-05T06:50:40.202367Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T06:50:40.202367Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Contrastive Tuning: A Little Help to Make Masked Autoencoders Forget","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CV","authors_text":"Andreas F\\\"urst, Benedikt Alkin, Elisabeth Rumetshofer, Johannes Lehner, Lukas Miklautz, Sepp Hochreiter","submitted_at":"2023-04-20T17:51:09Z","abstract_excerpt":"Masked Image Modeling (MIM) methods, like Masked Autoencoders (MAE), efficiently learn a rich representation of the input. However, for adapting to downstream tasks, they require a sufficient amount of labeled data since their rich features code not only objects but also less relevant image background. In contrast, Instance Discrimination (ID) methods focus on objects. In this work, we study how to combine the efficiency and scalability of MIM with the ability of ID to perform downstream classification in the absence of large amounts of labeled data. To this end, we introduce Masked Autoencode"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2304.10520","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/2304.10520/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":"2304.10520","created_at":"2026-07-05T06:50:40.202424+00:00"},{"alias_kind":"arxiv_version","alias_value":"2304.10520v2","created_at":"2026-07-05T06:50:40.202424+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2304.10520","created_at":"2026-07-05T06:50:40.202424+00:00"},{"alias_kind":"pith_short_12","alias_value":"JY7C5FA46RMW","created_at":"2026-07-05T06:50:40.202424+00:00"},{"alias_kind":"pith_short_16","alias_value":"JY7C5FA46RMWK4V7","created_at":"2026-07-05T06:50:40.202424+00:00"},{"alias_kind":"pith_short_8","alias_value":"JY7C5FA4","created_at":"2026-07-05T06:50:40.202424+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2502.06314","citing_title":"From Pixels to Components: Eigenvector Masking for Visual Representation Learning","ref_index":16,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/JY7C5FA46RMWK4V7PJDQH3WHC4","json":"https://pith.science/pith/JY7C5FA46RMWK4V7PJDQH3WHC4.json","graph_json":"https://pith.science/api/pith-number/JY7C5FA46RMWK4V7PJDQH3WHC4/graph.json","events_json":"https://pith.science/api/pith-number/JY7C5FA46RMWK4V7PJDQH3WHC4/events.json","paper":"https://pith.science/paper/JY7C5FA4"},"agent_actions":{"view_html":"https://pith.science/pith/JY7C5FA46RMWK4V7PJDQH3WHC4","download_json":"https://pith.science/pith/JY7C5FA46RMWK4V7PJDQH3WHC4.json","view_paper":"https://pith.science/paper/JY7C5FA4","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2304.10520&json=true","fetch_graph":"https://pith.science/api/pith-number/JY7C5FA46RMWK4V7PJDQH3WHC4/graph.json","fetch_events":"https://pith.science/api/pith-number/JY7C5FA46RMWK4V7PJDQH3WHC4/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/JY7C5FA46RMWK4V7PJDQH3WHC4/action/timestamp_anchor","attest_storage":"https://pith.science/pith/JY7C5FA46RMWK4V7PJDQH3WHC4/action/storage_attestation","attest_author":"https://pith.science/pith/JY7C5FA46RMWK4V7PJDQH3WHC4/action/author_attestation","sign_citation":"https://pith.science/pith/JY7C5FA46RMWK4V7PJDQH3WHC4/action/citation_signature","submit_replication":"https://pith.science/pith/JY7C5FA46RMWK4V7PJDQH3WHC4/action/replication_record"}},"created_at":"2026-07-05T06:50:40.202424+00:00","updated_at":"2026-07-05T06:50:40.202424+00:00"}