{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:LZPSKGDBV52RAFKRSRYRAIJFN3","short_pith_number":"pith:LZPSKGDB","schema_version":"1.0","canonical_sha256":"5e5f251861af7510155194711021256ecca1f622191c5bb0962a81ac2c4f9d5f","source":{"kind":"arxiv","id":"2310.13807","version":2},"attestation_state":"computed","paper":{"title":"Learning to (Learn at Test Time)","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Carlos Guestrin, Chloe Hsu, Karan Dalal, Sanmi Koyejo, Tatsunori Hashimoto, Xiaolong Wang, Xinhao Li, Xinlei Chen, Yu Sun","submitted_at":"2023-10-20T20:42:00Z","abstract_excerpt":"We reformulate the problem of supervised learning as learning to learn with two nested loops (i.e. learning problems). The inner loop learns on each individual instance with self-supervision before final prediction. The outer loop learns the self-supervised task used by the inner loop, such that its final prediction improves. Our inner loop turns out to be equivalent to linear attention when the inner-loop learner is only a linear model, and to self-attention when it is a kernel estimator. For practical comparison with linear or self-attention layers, we replace each of them in a transformer 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":"2310.13807","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2023-10-20T20:42:00Z","cross_cats_sorted":[],"title_canon_sha256":"7faab91cb8404a03f6b179d0622495fde0e7acb9771ee934c163d9b67afd4ffc","abstract_canon_sha256":"8619f36aeef4f078e0d6fd480dc65b66456d9f8c92c8c6f607905f60cc5612f0"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T07:30:55.795923Z","signature_b64":"5V5Wk9gx/eAYqRgEsXxDH6CFTiwYSqV5XdRA4XLvMHsDO0WwyFIdmv8ctaSNb0vt6ppjqa22X5AaLXW83U40AQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"5e5f251861af7510155194711021256ecca1f622191c5bb0962a81ac2c4f9d5f","last_reissued_at":"2026-07-05T07:30:55.795435Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T07:30:55.795435Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Learning to (Learn at Test Time)","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Carlos Guestrin, Chloe Hsu, Karan Dalal, Sanmi Koyejo, Tatsunori Hashimoto, Xiaolong Wang, Xinhao Li, Xinlei Chen, Yu Sun","submitted_at":"2023-10-20T20:42:00Z","abstract_excerpt":"We reformulate the problem of supervised learning as learning to learn with two nested loops (i.e. learning problems). The inner loop learns on each individual instance with self-supervision before final prediction. The outer loop learns the self-supervised task used by the inner loop, such that its final prediction improves. Our inner loop turns out to be equivalent to linear attention when the inner-loop learner is only a linear model, and to self-attention when it is a kernel estimator. For practical comparison with linear or self-attention layers, we replace each of them in a transformer w"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2310.13807","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/2310.13807/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":"2310.13807","created_at":"2026-07-05T07:30:55.795493+00:00"},{"alias_kind":"arxiv_version","alias_value":"2310.13807v2","created_at":"2026-07-05T07:30:55.795493+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2310.13807","created_at":"2026-07-05T07:30:55.795493+00:00"},{"alias_kind":"pith_short_12","alias_value":"LZPSKGDBV52R","created_at":"2026-07-05T07:30:55.795493+00:00"},{"alias_kind":"pith_short_16","alias_value":"LZPSKGDBV52RAFKR","created_at":"2026-07-05T07:30:55.795493+00:00"},{"alias_kind":"pith_short_8","alias_value":"LZPSKGDB","created_at":"2026-07-05T07:30:55.795493+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":6,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.11437","citing_title":"The Power of Test-Time Training for Approximate Sampling","ref_index":16,"is_internal_anchor":false},{"citing_arxiv_id":"2605.16981","citing_title":"Rethinking the State Update Gate for Long-Sequence Recurrent 3D Reconstruction","ref_index":19,"is_internal_anchor":false},{"citing_arxiv_id":"2601.16175","citing_title":"Learning to Discover at Test Time","ref_index":67,"is_internal_anchor":false},{"citing_arxiv_id":"2407.04620","citing_title":"Learning to (Learn at Test Time): RNNs with Expressive Hidden States","ref_index":70,"is_internal_anchor":false},{"citing_arxiv_id":"2406.07496","citing_title":"TextGrad: Automatic \"Differentiation\" via Text","ref_index":43,"is_internal_anchor":false},{"citing_arxiv_id":"2604.21100","citing_title":"Preconditioned DeltaNet: Curvature-aware Sequence Modeling for Linear Recurrences","ref_index":50,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/LZPSKGDBV52RAFKRSRYRAIJFN3","json":"https://pith.science/pith/LZPSKGDBV52RAFKRSRYRAIJFN3.json","graph_json":"https://pith.science/api/pith-number/LZPSKGDBV52RAFKRSRYRAIJFN3/graph.json","events_json":"https://pith.science/api/pith-number/LZPSKGDBV52RAFKRSRYRAIJFN3/events.json","paper":"https://pith.science/paper/LZPSKGDB"},"agent_actions":{"view_html":"https://pith.science/pith/LZPSKGDBV52RAFKRSRYRAIJFN3","download_json":"https://pith.science/pith/LZPSKGDBV52RAFKRSRYRAIJFN3.json","view_paper":"https://pith.science/paper/LZPSKGDB","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2310.13807&json=true","fetch_graph":"https://pith.science/api/pith-number/LZPSKGDBV52RAFKRSRYRAIJFN3/graph.json","fetch_events":"https://pith.science/api/pith-number/LZPSKGDBV52RAFKRSRYRAIJFN3/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/LZPSKGDBV52RAFKRSRYRAIJFN3/action/timestamp_anchor","attest_storage":"https://pith.science/pith/LZPSKGDBV52RAFKRSRYRAIJFN3/action/storage_attestation","attest_author":"https://pith.science/pith/LZPSKGDBV52RAFKRSRYRAIJFN3/action/author_attestation","sign_citation":"https://pith.science/pith/LZPSKGDBV52RAFKRSRYRAIJFN3/action/citation_signature","submit_replication":"https://pith.science/pith/LZPSKGDBV52RAFKRSRYRAIJFN3/action/replication_record"}},"created_at":"2026-07-05T07:30:55.795493+00:00","updated_at":"2026-07-05T07:30:55.795493+00:00"}