{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:YXHE2UREJILPJUZ25YWSNFSY45","short_pith_number":"pith:YXHE2URE","schema_version":"1.0","canonical_sha256":"c5ce4d52244a16f4d33aee2d269658e777c7a023ebffa39d41d14c3b222ac1c8","source":{"kind":"arxiv","id":"2304.06600","version":1},"attestation_state":"computed","paper":{"title":"Lossless Adaptation of Pretrained Vision Models For Robotic Manipulation","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CV","cs.RO"],"primary_cat":"cs.LG","authors_text":"Claudio Fantacci, Jon Scholz, Mohit Sharma, Nicolas Heess, Skanda Koppula, Yusuf Aytar, Yuxiang Zhou","submitted_at":"2023-04-13T15:06:28Z","abstract_excerpt":"Recent works have shown that large models pretrained on common visual learning tasks can provide useful representations for a wide range of specialized perception problems, as well as a variety of robotic manipulation tasks. While prior work on robotic manipulation has predominantly used frozen pretrained features, we demonstrate that in robotics this approach can fail to reach optimal performance, and that fine-tuning of the full model can lead to significantly better results. Unfortunately, fine-tuning disrupts the pretrained visual representation, and causes representational drift towards t"},"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":"2304.06600","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2023-04-13T15:06:28Z","cross_cats_sorted":["cs.CV","cs.RO"],"title_canon_sha256":"129671e44f934999607a8bc580f641357a9c0c6f9dc9874156c7ec4cd7f10f83","abstract_canon_sha256":"062e8d8083679f19979363d8fbee9ca71fad1dadc839448dcd3771afc378d841"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T06:00:45.615410Z","signature_b64":"r1iaGuhRCqpgichwr06IzJfRfgLx7FuRnzN9X8glagny2qaNSyL3+GmZujIfhpY2gGdzrn20JCWAwjzUtliiBw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"c5ce4d52244a16f4d33aee2d269658e777c7a023ebffa39d41d14c3b222ac1c8","last_reissued_at":"2026-07-05T06:00:45.615020Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T06:00:45.615020Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Lossless Adaptation of Pretrained Vision Models For Robotic Manipulation","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CV","cs.RO"],"primary_cat":"cs.LG","authors_text":"Claudio Fantacci, Jon Scholz, Mohit Sharma, Nicolas Heess, Skanda Koppula, Yusuf Aytar, Yuxiang Zhou","submitted_at":"2023-04-13T15:06:28Z","abstract_excerpt":"Recent works have shown that large models pretrained on common visual learning tasks can provide useful representations for a wide range of specialized perception problems, as well as a variety of robotic manipulation tasks. While prior work on robotic manipulation has predominantly used frozen pretrained features, we demonstrate that in robotics this approach can fail to reach optimal performance, and that fine-tuning of the full model can lead to significantly better results. Unfortunately, fine-tuning disrupts the pretrained visual representation, and causes representational drift towards t"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2304.06600","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/2304.06600/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":"2304.06600","created_at":"2026-07-05T06:00:45.615074+00:00"},{"alias_kind":"arxiv_version","alias_value":"2304.06600v1","created_at":"2026-07-05T06:00:45.615074+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2304.06600","created_at":"2026-07-05T06:00:45.615074+00:00"},{"alias_kind":"pith_short_12","alias_value":"YXHE2UREJILP","created_at":"2026-07-05T06:00:45.615074+00:00"},{"alias_kind":"pith_short_16","alias_value":"YXHE2UREJILPJUZ2","created_at":"2026-07-05T06:00:45.615074+00:00"},{"alias_kind":"pith_short_8","alias_value":"YXHE2URE","created_at":"2026-07-05T06:00:45.615074+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2409.16283","citing_title":"Gen2Act: Human Video Generation in Novel Scenarios enables Generalizable Robot Manipulation","ref_index":37,"is_internal_anchor":false},{"citing_arxiv_id":"2604.23121","citing_title":"Breaking Lock-In: Preserving Steerability under Low-Data VLA Post-Training","ref_index":44,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/YXHE2UREJILPJUZ25YWSNFSY45","json":"https://pith.science/pith/YXHE2UREJILPJUZ25YWSNFSY45.json","graph_json":"https://pith.science/api/pith-number/YXHE2UREJILPJUZ25YWSNFSY45/graph.json","events_json":"https://pith.science/api/pith-number/YXHE2UREJILPJUZ25YWSNFSY45/events.json","paper":"https://pith.science/paper/YXHE2URE"},"agent_actions":{"view_html":"https://pith.science/pith/YXHE2UREJILPJUZ25YWSNFSY45","download_json":"https://pith.science/pith/YXHE2UREJILPJUZ25YWSNFSY45.json","view_paper":"https://pith.science/paper/YXHE2URE","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2304.06600&json=true","fetch_graph":"https://pith.science/api/pith-number/YXHE2UREJILPJUZ25YWSNFSY45/graph.json","fetch_events":"https://pith.science/api/pith-number/YXHE2UREJILPJUZ25YWSNFSY45/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/YXHE2UREJILPJUZ25YWSNFSY45/action/timestamp_anchor","attest_storage":"https://pith.science/pith/YXHE2UREJILPJUZ25YWSNFSY45/action/storage_attestation","attest_author":"https://pith.science/pith/YXHE2UREJILPJUZ25YWSNFSY45/action/author_attestation","sign_citation":"https://pith.science/pith/YXHE2UREJILPJUZ25YWSNFSY45/action/citation_signature","submit_replication":"https://pith.science/pith/YXHE2UREJILPJUZ25YWSNFSY45/action/replication_record"}},"created_at":"2026-07-05T06:00:45.615074+00:00","updated_at":"2026-07-05T06:00:45.615074+00:00"}