{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2020:6Y2LQRM5KDLWDZFH5JQH4ETDJG","short_pith_number":"pith:6Y2LQRM5","schema_version":"1.0","canonical_sha256":"f634b8459d50d761e4a7ea607e126349af9424277c771f04c8ac4dd8d6c8e6cf","source":{"kind":"arxiv","id":"2003.11942","version":3},"attestation_state":"computed","paper":{"title":"Towards Backward-Compatible Representation Learning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Stefano Soatto, Wei Xia, Yantao Shen, Yuanjun Xiong","submitted_at":"2020-03-26T14:34:09Z","abstract_excerpt":"We propose a way to learn visual features that are compatible with previously computed ones even when they have different dimensions and are learned via different neural network architectures and loss functions. Compatible means that, if such features are used to compare images, then \"new\" features can be compared directly to \"old\" features, so they can be used interchangeably. This enables visual search systems to bypass computing new features for all previously seen images when updating the embedding models, a process known as backfilling. Backward compatibility is critical to quickly deploy"},"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":"2003.11942","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2020-03-26T14:34:09Z","cross_cats_sorted":[],"title_canon_sha256":"4def1ec52dad89031150511925d7998076d30e2ba5de1cf27e6a18e34ad3a197","abstract_canon_sha256":"b600efae778cd51029a4f5f8f939457d8373520f61c9b7fa6b7a08030adaf082"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T02:04:55.753497Z","signature_b64":"S721CXqJ1NqlJ+1GU72zTvaGyGgs1uu7C+YlSe2BuMq942yZapNw3RlXNTHyKdC3e0mz92dIXLs/OLw4ZftODQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"f634b8459d50d761e4a7ea607e126349af9424277c771f04c8ac4dd8d6c8e6cf","last_reissued_at":"2026-07-05T02:04:55.752988Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T02:04:55.752988Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Towards Backward-Compatible Representation Learning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Stefano Soatto, Wei Xia, Yantao Shen, Yuanjun Xiong","submitted_at":"2020-03-26T14:34:09Z","abstract_excerpt":"We propose a way to learn visual features that are compatible with previously computed ones even when they have different dimensions and are learned via different neural network architectures and loss functions. Compatible means that, if such features are used to compare images, then \"new\" features can be compared directly to \"old\" features, so they can be used interchangeably. This enables visual search systems to bypass computing new features for all previously seen images when updating the embedding models, a process known as backfilling. Backward compatibility is critical to quickly deploy"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2003.11942","kind":"arxiv","version":3},"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/2003.11942/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":"2003.11942","created_at":"2026-07-05T02:04:55.753047+00:00"},{"alias_kind":"arxiv_version","alias_value":"2003.11942v3","created_at":"2026-07-05T02:04:55.753047+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2003.11942","created_at":"2026-07-05T02:04:55.753047+00:00"},{"alias_kind":"pith_short_12","alias_value":"6Y2LQRM5KDLW","created_at":"2026-07-05T02:04:55.753047+00:00"},{"alias_kind":"pith_short_16","alias_value":"6Y2LQRM5KDLWDZFH","created_at":"2026-07-05T02:04:55.753047+00:00"},{"alias_kind":"pith_short_8","alias_value":"6Y2LQRM5","created_at":"2026-07-05T02:04:55.753047+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.31100","citing_title":"Vector Linking via Cross-Model Local Isometric Consistency","ref_index":23,"is_internal_anchor":false},{"citing_arxiv_id":"2205.01068","citing_title":"OPT: Open Pre-trained Transformer Language Models","ref_index":8,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/6Y2LQRM5KDLWDZFH5JQH4ETDJG","json":"https://pith.science/pith/6Y2LQRM5KDLWDZFH5JQH4ETDJG.json","graph_json":"https://pith.science/api/pith-number/6Y2LQRM5KDLWDZFH5JQH4ETDJG/graph.json","events_json":"https://pith.science/api/pith-number/6Y2LQRM5KDLWDZFH5JQH4ETDJG/events.json","paper":"https://pith.science/paper/6Y2LQRM5"},"agent_actions":{"view_html":"https://pith.science/pith/6Y2LQRM5KDLWDZFH5JQH4ETDJG","download_json":"https://pith.science/pith/6Y2LQRM5KDLWDZFH5JQH4ETDJG.json","view_paper":"https://pith.science/paper/6Y2LQRM5","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2003.11942&json=true","fetch_graph":"https://pith.science/api/pith-number/6Y2LQRM5KDLWDZFH5JQH4ETDJG/graph.json","fetch_events":"https://pith.science/api/pith-number/6Y2LQRM5KDLWDZFH5JQH4ETDJG/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/6Y2LQRM5KDLWDZFH5JQH4ETDJG/action/timestamp_anchor","attest_storage":"https://pith.science/pith/6Y2LQRM5KDLWDZFH5JQH4ETDJG/action/storage_attestation","attest_author":"https://pith.science/pith/6Y2LQRM5KDLWDZFH5JQH4ETDJG/action/author_attestation","sign_citation":"https://pith.science/pith/6Y2LQRM5KDLWDZFH5JQH4ETDJG/action/citation_signature","submit_replication":"https://pith.science/pith/6Y2LQRM5KDLWDZFH5JQH4ETDJG/action/replication_record"}},"created_at":"2026-07-05T02:04:55.753047+00:00","updated_at":"2026-07-05T02:04:55.753047+00:00"}