{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2020:XSXZJ7MLSVP6UBMMPLM56MOPFV","short_pith_number":"pith:XSXZJ7ML","schema_version":"1.0","canonical_sha256":"bcaf94fd8b955fea058c7ad9df31cf2d50a34d7d4e977da18e6a017ab4d2a8fd","source":{"kind":"arxiv","id":"2010.15775","version":3},"attestation_state":"computed","paper":{"title":"Understanding the Failure Modes of Out-of-Distribution Generalization","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CV","stat.ML"],"primary_cat":"cs.LG","authors_text":"Anders Andreassen, Behnam Neyshabur, Vaishnavh Nagarajan","submitted_at":"2020-10-29T17:19:03Z","abstract_excerpt":"Empirical studies suggest that machine learning models often rely on features, such as the background, that may be spuriously correlated with the label only during training time, resulting in poor accuracy during test-time. In this work, we identify the fundamental factors that give rise to this behavior, by explaining why models fail this way {\\em even} in easy-to-learn tasks where one would expect these models to succeed. In particular, through a theoretical study of gradient-descent-trained linear classifiers on some easy-to-learn tasks, we uncover two complementary failure modes. These mod"},"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":"2010.15775","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2020-10-29T17:19:03Z","cross_cats_sorted":["cs.CV","stat.ML"],"title_canon_sha256":"a610302533d17dfb200abb73bf97d9d0d4bd5239652a5b60c49cd165fce4fcd2","abstract_canon_sha256":"7e1886ce0632c949788945ef6479622be097a86789cbc3a2cacdb5f8909e6530"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:04:06.219572Z","signature_b64":"TCsJMCv0fX3gOL/JpF3ZUSEua/Rhn17v7DvTFve1oeRw8FHxemuiZe6y6AqFysJgOQPZNhQVxoxbEGKZ3kQ7Dw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"bcaf94fd8b955fea058c7ad9df31cf2d50a34d7d4e977da18e6a017ab4d2a8fd","last_reissued_at":"2026-07-05T09:04:06.219176Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:04:06.219176Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Understanding the Failure Modes of Out-of-Distribution Generalization","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CV","stat.ML"],"primary_cat":"cs.LG","authors_text":"Anders Andreassen, Behnam Neyshabur, Vaishnavh Nagarajan","submitted_at":"2020-10-29T17:19:03Z","abstract_excerpt":"Empirical studies suggest that machine learning models often rely on features, such as the background, that may be spuriously correlated with the label only during training time, resulting in poor accuracy during test-time. In this work, we identify the fundamental factors that give rise to this behavior, by explaining why models fail this way {\\em even} in easy-to-learn tasks where one would expect these models to succeed. In particular, through a theoretical study of gradient-descent-trained linear classifiers on some easy-to-learn tasks, we uncover two complementary failure modes. These mod"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2010.15775","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/2010.15775/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":"2010.15775","created_at":"2026-07-05T09:04:06.219231+00:00"},{"alias_kind":"arxiv_version","alias_value":"2010.15775v3","created_at":"2026-07-05T09:04:06.219231+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2010.15775","created_at":"2026-07-05T09:04:06.219231+00:00"},{"alias_kind":"pith_short_12","alias_value":"XSXZJ7MLSVP6","created_at":"2026-07-05T09:04:06.219231+00:00"},{"alias_kind":"pith_short_16","alias_value":"XSXZJ7MLSVP6UBMM","created_at":"2026-07-05T09:04:06.219231+00:00"},{"alias_kind":"pith_short_8","alias_value":"XSXZJ7ML","created_at":"2026-07-05T09:04:06.219231+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.20771","citing_title":"Cumulative Meta-Learning from Active Learning Queries for Robustness to Spurious Correlations","ref_index":130,"is_internal_anchor":false},{"citing_arxiv_id":"2406.10162","citing_title":"Sycophancy to Subterfuge: Investigating Reward-Tampering in Large Language Models","ref_index":63,"is_internal_anchor":false},{"citing_arxiv_id":"2602.05353","citing_title":"AgentXRay: White-Boxing Agentic Systems via Workflow Reconstruction","ref_index":12,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/XSXZJ7MLSVP6UBMMPLM56MOPFV","json":"https://pith.science/pith/XSXZJ7MLSVP6UBMMPLM56MOPFV.json","graph_json":"https://pith.science/api/pith-number/XSXZJ7MLSVP6UBMMPLM56MOPFV/graph.json","events_json":"https://pith.science/api/pith-number/XSXZJ7MLSVP6UBMMPLM56MOPFV/events.json","paper":"https://pith.science/paper/XSXZJ7ML"},"agent_actions":{"view_html":"https://pith.science/pith/XSXZJ7MLSVP6UBMMPLM56MOPFV","download_json":"https://pith.science/pith/XSXZJ7MLSVP6UBMMPLM56MOPFV.json","view_paper":"https://pith.science/paper/XSXZJ7ML","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2010.15775&json=true","fetch_graph":"https://pith.science/api/pith-number/XSXZJ7MLSVP6UBMMPLM56MOPFV/graph.json","fetch_events":"https://pith.science/api/pith-number/XSXZJ7MLSVP6UBMMPLM56MOPFV/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/XSXZJ7MLSVP6UBMMPLM56MOPFV/action/timestamp_anchor","attest_storage":"https://pith.science/pith/XSXZJ7MLSVP6UBMMPLM56MOPFV/action/storage_attestation","attest_author":"https://pith.science/pith/XSXZJ7MLSVP6UBMMPLM56MOPFV/action/author_attestation","sign_citation":"https://pith.science/pith/XSXZJ7MLSVP6UBMMPLM56MOPFV/action/citation_signature","submit_replication":"https://pith.science/pith/XSXZJ7MLSVP6UBMMPLM56MOPFV/action/replication_record"}},"created_at":"2026-07-05T09:04:06.219231+00:00","updated_at":"2026-07-05T09:04:06.219231+00:00"}