{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2021:BYDNI2XOFCKVL7P33FSTZJNKHG","short_pith_number":"pith:BYDNI2XO","schema_version":"1.0","canonical_sha256":"0e06d46aee289555fdfbd9653ca5aa39ab3081b9bfc66f2c01a434b32633620c","source":{"kind":"arxiv","id":"2109.02934","version":3},"attestation_state":"computed","paper":{"title":"Fishr: Invariant Gradient Variances for Out-of-Distribution Generalization","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CV"],"primary_cat":"cs.LG","authors_text":"Alexandre Rame, Corentin Dancette, Matthieu Cord","submitted_at":"2021-09-07T08:36:09Z","abstract_excerpt":"Learning robust models that generalize well under changes in the data distribution is critical for real-world applications. To this end, there has been a growing surge of interest to learn simultaneously from multiple training domains - while enforcing different types of invariance across those domains. Yet, all existing approaches fail to show systematic benefits under controlled evaluation protocols. In this paper, we introduce a new regularization - named Fishr - that enforces domain invariance in the space of the gradients of the loss: specifically, the domain-level variances of gradients "},"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":"2109.02934","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2021-09-07T08:36:09Z","cross_cats_sorted":["cs.AI","cs.CV"],"title_canon_sha256":"cc5189c31621dbb062c1ba830ea650e038c315c5672c4097eee90a4220ab59a8","abstract_canon_sha256":"06dc088b345662bf2ad0be2866e03be881cece9d00646cdf8a86a4a6da28c244"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T04:28:07.674883Z","signature_b64":"p3dH9SpxxnMw3iqytNgBmnlNF/l+3574Qoa8RpxLFKD4Iee23KVxi61sFDfcca/ukGG1N88M40k8MzEJm58WAg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"0e06d46aee289555fdfbd9653ca5aa39ab3081b9bfc66f2c01a434b32633620c","last_reissued_at":"2026-07-05T04:28:07.674379Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T04:28:07.674379Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Fishr: Invariant Gradient Variances for Out-of-Distribution Generalization","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CV"],"primary_cat":"cs.LG","authors_text":"Alexandre Rame, Corentin Dancette, Matthieu Cord","submitted_at":"2021-09-07T08:36:09Z","abstract_excerpt":"Learning robust models that generalize well under changes in the data distribution is critical for real-world applications. To this end, there has been a growing surge of interest to learn simultaneously from multiple training domains - while enforcing different types of invariance across those domains. Yet, all existing approaches fail to show systematic benefits under controlled evaluation protocols. In this paper, we introduce a new regularization - named Fishr - that enforces domain invariance in the space of the gradients of the loss: specifically, the domain-level variances of gradients "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2109.02934","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/2109.02934/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":"2109.02934","created_at":"2026-07-05T04:28:07.674441+00:00"},{"alias_kind":"arxiv_version","alias_value":"2109.02934v3","created_at":"2026-07-05T04:28:07.674441+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2109.02934","created_at":"2026-07-05T04:28:07.674441+00:00"},{"alias_kind":"pith_short_12","alias_value":"BYDNI2XOFCKV","created_at":"2026-07-05T04:28:07.674441+00:00"},{"alias_kind":"pith_short_16","alias_value":"BYDNI2XOFCKVL7P3","created_at":"2026-07-05T04:28:07.674441+00:00"},{"alias_kind":"pith_short_8","alias_value":"BYDNI2XO","created_at":"2026-07-05T04:28:07.674441+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2209.14742","citing_title":"Learning Gradient-based Mixup with Extrapolation toward Flatter Minima for Domain Generalization","ref_index":34,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/BYDNI2XOFCKVL7P33FSTZJNKHG","json":"https://pith.science/pith/BYDNI2XOFCKVL7P33FSTZJNKHG.json","graph_json":"https://pith.science/api/pith-number/BYDNI2XOFCKVL7P33FSTZJNKHG/graph.json","events_json":"https://pith.science/api/pith-number/BYDNI2XOFCKVL7P33FSTZJNKHG/events.json","paper":"https://pith.science/paper/BYDNI2XO"},"agent_actions":{"view_html":"https://pith.science/pith/BYDNI2XOFCKVL7P33FSTZJNKHG","download_json":"https://pith.science/pith/BYDNI2XOFCKVL7P33FSTZJNKHG.json","view_paper":"https://pith.science/paper/BYDNI2XO","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2109.02934&json=true","fetch_graph":"https://pith.science/api/pith-number/BYDNI2XOFCKVL7P33FSTZJNKHG/graph.json","fetch_events":"https://pith.science/api/pith-number/BYDNI2XOFCKVL7P33FSTZJNKHG/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/BYDNI2XOFCKVL7P33FSTZJNKHG/action/timestamp_anchor","attest_storage":"https://pith.science/pith/BYDNI2XOFCKVL7P33FSTZJNKHG/action/storage_attestation","attest_author":"https://pith.science/pith/BYDNI2XOFCKVL7P33FSTZJNKHG/action/author_attestation","sign_citation":"https://pith.science/pith/BYDNI2XOFCKVL7P33FSTZJNKHG/action/citation_signature","submit_replication":"https://pith.science/pith/BYDNI2XOFCKVL7P33FSTZJNKHG/action/replication_record"}},"created_at":"2026-07-05T04:28:07.674441+00:00","updated_at":"2026-07-05T04:28:07.674441+00:00"}