{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:FL7ESOVRTRN5QYYZ6FNLJ7KSA6","short_pith_number":"pith:FL7ESOVR","schema_version":"1.0","canonical_sha256":"2afe493ab19c5bd86319f15ab4fd52079f18b96e8e54877b6c7812c8c0417ca2","source":{"kind":"arxiv","id":"2402.10093","version":4},"attestation_state":"computed","paper":{"title":"MIM-Refiner: A Contrastive Learning Boost from Intermediate Pre-Trained Representations","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CV","authors_text":"Benedikt Alkin, Johannes Brandstetter, Lukas Miklautz, Sepp Hochreiter","submitted_at":"2024-02-15T16:46:16Z","abstract_excerpt":"We introduce MIM (Masked Image Modeling)-Refiner, a contrastive learning boost for pre-trained MIM models. MIM-Refiner is motivated by the insight that strong representations within MIM models generally reside in intermediate layers. Accordingly, MIM-Refiner leverages multiple contrastive heads that are connected to different intermediate layers. In each head, a modified nearest neighbor objective constructs semantic clusters that capture semantic information which improves performance on downstream tasks, including off-the-shelf and fine-tuning settings.\n  The refinement process is short and "},"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":"2402.10093","kind":"arxiv","version":4},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2024-02-15T16:46:16Z","cross_cats_sorted":["cs.AI","cs.LG"],"title_canon_sha256":"3e70f58284d0fdc42f5e6eefba4b0c9db6685e4677dbaff99dfe36c7ab7574b5","abstract_canon_sha256":"ab4691979d4e3ce7b9ba323906c0d7cf32c1c5a787a6e14ca40a2bc1c5d0b8c3"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:17:46.217772Z","signature_b64":"Mq9gv7NfqTCBDreB/STEl47yY4Kn4h0kddQ065tJMxyTxDdK3ec6L0WM6VoFCU2JX2mPacSoYbOrqC96JBU8Aw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"2afe493ab19c5bd86319f15ab4fd52079f18b96e8e54877b6c7812c8c0417ca2","last_reissued_at":"2026-07-05T10:17:46.217297Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:17:46.217297Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"MIM-Refiner: A Contrastive Learning Boost from Intermediate Pre-Trained Representations","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CV","authors_text":"Benedikt Alkin, Johannes Brandstetter, Lukas Miklautz, Sepp Hochreiter","submitted_at":"2024-02-15T16:46:16Z","abstract_excerpt":"We introduce MIM (Masked Image Modeling)-Refiner, a contrastive learning boost for pre-trained MIM models. MIM-Refiner is motivated by the insight that strong representations within MIM models generally reside in intermediate layers. Accordingly, MIM-Refiner leverages multiple contrastive heads that are connected to different intermediate layers. In each head, a modified nearest neighbor objective constructs semantic clusters that capture semantic information which improves performance on downstream tasks, including off-the-shelf and fine-tuning settings.\n  The refinement process is short and "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2402.10093","kind":"arxiv","version":4},"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/2402.10093/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":"2402.10093","created_at":"2026-07-05T10:17:46.217362+00:00"},{"alias_kind":"arxiv_version","alias_value":"2402.10093v4","created_at":"2026-07-05T10:17:46.217362+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2402.10093","created_at":"2026-07-05T10:17:46.217362+00:00"},{"alias_kind":"pith_short_12","alias_value":"FL7ESOVRTRN5","created_at":"2026-07-05T10:17:46.217362+00:00"},{"alias_kind":"pith_short_16","alias_value":"FL7ESOVRTRN5QYYZ","created_at":"2026-07-05T10:17:46.217362+00:00"},{"alias_kind":"pith_short_8","alias_value":"FL7ESOVR","created_at":"2026-07-05T10:17:46.217362+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2412.16247","citing_title":"Towards scientific discovery with dictionary learning: Extracting biological concepts from microscopy foundation models","ref_index":2006,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/FL7ESOVRTRN5QYYZ6FNLJ7KSA6","json":"https://pith.science/pith/FL7ESOVRTRN5QYYZ6FNLJ7KSA6.json","graph_json":"https://pith.science/api/pith-number/FL7ESOVRTRN5QYYZ6FNLJ7KSA6/graph.json","events_json":"https://pith.science/api/pith-number/FL7ESOVRTRN5QYYZ6FNLJ7KSA6/events.json","paper":"https://pith.science/paper/FL7ESOVR"},"agent_actions":{"view_html":"https://pith.science/pith/FL7ESOVRTRN5QYYZ6FNLJ7KSA6","download_json":"https://pith.science/pith/FL7ESOVRTRN5QYYZ6FNLJ7KSA6.json","view_paper":"https://pith.science/paper/FL7ESOVR","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2402.10093&json=true","fetch_graph":"https://pith.science/api/pith-number/FL7ESOVRTRN5QYYZ6FNLJ7KSA6/graph.json","fetch_events":"https://pith.science/api/pith-number/FL7ESOVRTRN5QYYZ6FNLJ7KSA6/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/FL7ESOVRTRN5QYYZ6FNLJ7KSA6/action/timestamp_anchor","attest_storage":"https://pith.science/pith/FL7ESOVRTRN5QYYZ6FNLJ7KSA6/action/storage_attestation","attest_author":"https://pith.science/pith/FL7ESOVRTRN5QYYZ6FNLJ7KSA6/action/author_attestation","sign_citation":"https://pith.science/pith/FL7ESOVRTRN5QYYZ6FNLJ7KSA6/action/citation_signature","submit_replication":"https://pith.science/pith/FL7ESOVRTRN5QYYZ6FNLJ7KSA6/action/replication_record"}},"created_at":"2026-07-05T10:17:46.217362+00:00","updated_at":"2026-07-05T10:17:46.217362+00:00"}