{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2021:PIJOYBP6KVDP46QMQDVDECOVBF","short_pith_number":"pith:PIJOYBP6","schema_version":"1.0","canonical_sha256":"7a12ec05fe5546fe7a0c80ea3209d50962d740cb2a75d4662810f8ca657d3326","source":{"kind":"arxiv","id":"2110.06177","version":4},"attestation_state":"computed","paper":{"title":"Tracking the risk of a deployed model and detecting harmful distribution shifts","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"stat.ML","authors_text":"Aaditya Ramdas, Aleksandr Podkopaev","submitted_at":"2021-10-12T17:21:41Z","abstract_excerpt":"When deployed in the real world, machine learning models inevitably encounter changes in the data distribution, and certain -- but not all -- distribution shifts could result in significant performance degradation. In practice, it may make sense to ignore benign shifts, under which the performance of a deployed model does not degrade substantially, making interventions by a human expert (or model retraining) unnecessary. While several works have developed tests for distribution shifts, these typically either use non-sequential methods, or detect arbitrary shifts (benign or harmful), or both. W"},"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":"2110.06177","kind":"arxiv","version":4},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"stat.ML","submitted_at":"2021-10-12T17:21:41Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"dd9e127631a06ed960dd91db8751fb51ee16ffa4806d78b51dc8a942646a9cad","abstract_canon_sha256":"401f077b4a7d0ad1a973667e2818e303ae457d8d08019bc8462cd93aee573588"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T04:20:30.122109Z","signature_b64":"mEeuz2gZVhseHDE5XLGDcK9ri6Vd3IhTKOdOc62jEa9jhop9XrKdfdbu7jm3mJkG6b+xGKE7rYhbg1m6PjNpDA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"7a12ec05fe5546fe7a0c80ea3209d50962d740cb2a75d4662810f8ca657d3326","last_reissued_at":"2026-07-05T04:20:30.121630Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T04:20:30.121630Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Tracking the risk of a deployed model and detecting harmful distribution shifts","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"stat.ML","authors_text":"Aaditya Ramdas, Aleksandr Podkopaev","submitted_at":"2021-10-12T17:21:41Z","abstract_excerpt":"When deployed in the real world, machine learning models inevitably encounter changes in the data distribution, and certain -- but not all -- distribution shifts could result in significant performance degradation. In practice, it may make sense to ignore benign shifts, under which the performance of a deployed model does not degrade substantially, making interventions by a human expert (or model retraining) unnecessary. While several works have developed tests for distribution shifts, these typically either use non-sequential methods, or detect arbitrary shifts (benign or harmful), or both. W"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2110.06177","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/2110.06177/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":"2110.06177","created_at":"2026-07-05T04:20:30.121698+00:00"},{"alias_kind":"arxiv_version","alias_value":"2110.06177v4","created_at":"2026-07-05T04:20:30.121698+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2110.06177","created_at":"2026-07-05T04:20:30.121698+00:00"},{"alias_kind":"pith_short_12","alias_value":"PIJOYBP6KVDP","created_at":"2026-07-05T04:20:30.121698+00:00"},{"alias_kind":"pith_short_16","alias_value":"PIJOYBP6KVDP46QM","created_at":"2026-07-05T04:20:30.121698+00:00"},{"alias_kind":"pith_short_8","alias_value":"PIJOYBP6","created_at":"2026-07-05T04:20:30.121698+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":4,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2607.02510","citing_title":"Online Safety Monitoring for LLMs","ref_index":22,"is_internal_anchor":false},{"citing_arxiv_id":"2606.11949","citing_title":"Online Shift Detection and Conformal Adaptation for Deployed Safety Classifiers","ref_index":27,"is_internal_anchor":false},{"citing_arxiv_id":"2606.11949","citing_title":"Online Shift Detection and Conformal Adaptation for Deployed Safety Classifiers","ref_index":38,"is_internal_anchor":false},{"citing_arxiv_id":"2107.07511","citing_title":"A Gentle Introduction to Conformal Prediction and Distribution-Free Uncertainty Quantification","ref_index":108,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/PIJOYBP6KVDP46QMQDVDECOVBF","json":"https://pith.science/pith/PIJOYBP6KVDP46QMQDVDECOVBF.json","graph_json":"https://pith.science/api/pith-number/PIJOYBP6KVDP46QMQDVDECOVBF/graph.json","events_json":"https://pith.science/api/pith-number/PIJOYBP6KVDP46QMQDVDECOVBF/events.json","paper":"https://pith.science/paper/PIJOYBP6"},"agent_actions":{"view_html":"https://pith.science/pith/PIJOYBP6KVDP46QMQDVDECOVBF","download_json":"https://pith.science/pith/PIJOYBP6KVDP46QMQDVDECOVBF.json","view_paper":"https://pith.science/paper/PIJOYBP6","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2110.06177&json=true","fetch_graph":"https://pith.science/api/pith-number/PIJOYBP6KVDP46QMQDVDECOVBF/graph.json","fetch_events":"https://pith.science/api/pith-number/PIJOYBP6KVDP46QMQDVDECOVBF/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/PIJOYBP6KVDP46QMQDVDECOVBF/action/timestamp_anchor","attest_storage":"https://pith.science/pith/PIJOYBP6KVDP46QMQDVDECOVBF/action/storage_attestation","attest_author":"https://pith.science/pith/PIJOYBP6KVDP46QMQDVDECOVBF/action/author_attestation","sign_citation":"https://pith.science/pith/PIJOYBP6KVDP46QMQDVDECOVBF/action/citation_signature","submit_replication":"https://pith.science/pith/PIJOYBP6KVDP46QMQDVDECOVBF/action/replication_record"}},"created_at":"2026-07-05T04:20:30.121698+00:00","updated_at":"2026-07-05T04:20:30.121698+00:00"}