{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2022:67JQCB22U4P6KXH4PVX2RIRGSQ","short_pith_number":"pith:67JQCB22","schema_version":"1.0","canonical_sha256":"f7d301075aa71fe55cfc7d6fa8a22694392b9716d463f159e1b8fa563bc04217","source":{"kind":"arxiv","id":"2207.09640","version":2},"attestation_state":"computed","paper":{"title":"Test-Time Adaptation via Conjugate Pseudo-labels","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Aditi Raghunathan, Mingjie Sun, Sachin Goyal, Zico Kolter","submitted_at":"2022-07-20T04:02:19Z","abstract_excerpt":"Test-time adaptation (TTA) refers to adapting neural networks to distribution shifts, with access to only the unlabeled test samples from the new domain at test-time. Prior TTA methods optimize over unsupervised objectives such as the entropy of model predictions in TENT [Wang et al., 2021], but it is unclear what exactly makes a good TTA loss. In this paper, we start by presenting a surprising phenomenon: if we attempt to meta-learn the best possible TTA loss over a wide class of functions, then we recover a function that is remarkably similar to (a temperature-scaled version of) the softmax-"},"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":"2207.09640","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","primary_cat":"cs.LG","submitted_at":"2022-07-20T04:02:19Z","cross_cats_sorted":[],"title_canon_sha256":"fb44417bd5ae3ae82644d745be6c7d22d480dc221164263fb53afa059ef783d1","abstract_canon_sha256":"4bdddc2742a780ec2ddc8678739749cf0736aec6bfc86c067900cdfdea416a70"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T05:18:47.710990Z","signature_b64":"xB19fz0NcHdXn3QZKxSADYOUOrdhhJus/obkdBP2Bkr6HlfQ9W53m9s1pathV/WzaGZJWNccoUnL8YyV4rsjBg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"f7d301075aa71fe55cfc7d6fa8a22694392b9716d463f159e1b8fa563bc04217","last_reissued_at":"2026-07-05T05:18:47.710265Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T05:18:47.710265Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Test-Time Adaptation via Conjugate Pseudo-labels","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Aditi Raghunathan, Mingjie Sun, Sachin Goyal, Zico Kolter","submitted_at":"2022-07-20T04:02:19Z","abstract_excerpt":"Test-time adaptation (TTA) refers to adapting neural networks to distribution shifts, with access to only the unlabeled test samples from the new domain at test-time. Prior TTA methods optimize over unsupervised objectives such as the entropy of model predictions in TENT [Wang et al., 2021], but it is unclear what exactly makes a good TTA loss. In this paper, we start by presenting a surprising phenomenon: if we attempt to meta-learn the best possible TTA loss over a wide class of functions, then we recover a function that is remarkably similar to (a temperature-scaled version of) the softmax-"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2207.09640","kind":"arxiv","version":2},"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/2207.09640/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":"2207.09640","created_at":"2026-07-05T05:18:47.710332+00:00"},{"alias_kind":"arxiv_version","alias_value":"2207.09640v2","created_at":"2026-07-05T05:18:47.710332+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2207.09640","created_at":"2026-07-05T05:18:47.710332+00:00"},{"alias_kind":"pith_short_12","alias_value":"67JQCB22U4P6","created_at":"2026-07-05T05:18:47.710332+00:00"},{"alias_kind":"pith_short_16","alias_value":"67JQCB22U4P6KXH4","created_at":"2026-07-05T05:18:47.710332+00:00"},{"alias_kind":"pith_short_8","alias_value":"67JQCB22","created_at":"2026-07-05T05:18:47.710332+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2412.15511","citing_title":"RESQUE: Quantifying Estimator to Task and Distribution Shift for Sustainable Model Reusability","ref_index":21,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/67JQCB22U4P6KXH4PVX2RIRGSQ","json":"https://pith.science/pith/67JQCB22U4P6KXH4PVX2RIRGSQ.json","graph_json":"https://pith.science/api/pith-number/67JQCB22U4P6KXH4PVX2RIRGSQ/graph.json","events_json":"https://pith.science/api/pith-number/67JQCB22U4P6KXH4PVX2RIRGSQ/events.json","paper":"https://pith.science/paper/67JQCB22"},"agent_actions":{"view_html":"https://pith.science/pith/67JQCB22U4P6KXH4PVX2RIRGSQ","download_json":"https://pith.science/pith/67JQCB22U4P6KXH4PVX2RIRGSQ.json","view_paper":"https://pith.science/paper/67JQCB22","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2207.09640&json=true","fetch_graph":"https://pith.science/api/pith-number/67JQCB22U4P6KXH4PVX2RIRGSQ/graph.json","fetch_events":"https://pith.science/api/pith-number/67JQCB22U4P6KXH4PVX2RIRGSQ/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/67JQCB22U4P6KXH4PVX2RIRGSQ/action/timestamp_anchor","attest_storage":"https://pith.science/pith/67JQCB22U4P6KXH4PVX2RIRGSQ/action/storage_attestation","attest_author":"https://pith.science/pith/67JQCB22U4P6KXH4PVX2RIRGSQ/action/author_attestation","sign_citation":"https://pith.science/pith/67JQCB22U4P6KXH4PVX2RIRGSQ/action/citation_signature","submit_replication":"https://pith.science/pith/67JQCB22U4P6KXH4PVX2RIRGSQ/action/replication_record"}},"created_at":"2026-07-05T05:18:47.710332+00:00","updated_at":"2026-07-05T05:18:47.710332+00:00"}