{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2022:I5IS4VBMOATATTO2QNOLPVKG2P","short_pith_number":"pith:I5IS4VBM","schema_version":"1.0","canonical_sha256":"47512e542c702609cdda835cb7d546d3c4af46b17186ad0c9aa74c8ad4d86374","source":{"kind":"arxiv","id":"2212.07677","version":2},"attestation_state":"computed","paper":{"title":"Transformers learn in-context by gradient descent","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.LG","authors_text":"Alexander Mordvintsev, Andrey Zhmoginov, Ettore Randazzo, Eyvind Niklasson, Jo\\~ao Sacramento, Johannes von Oswald, Max Vladymyrov","submitted_at":"2022-12-15T09:21:21Z","abstract_excerpt":"At present, the mechanisms of in-context learning in Transformers are not well understood and remain mostly an intuition. In this paper, we suggest that training Transformers on auto-regressive objectives is closely related to gradient-based meta-learning formulations. We start by providing a simple weight construction that shows the equivalence of data transformations induced by 1) a single linear self-attention layer and by 2) gradient-descent (GD) on a regression loss. Motivated by that construction, we show empirically that when training self-attention-only Transformers on simple regressio"},"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":"2212.07677","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2022-12-15T09:21:21Z","cross_cats_sorted":["cs.AI","cs.CL"],"title_canon_sha256":"42a78aec72b26e85f668c09d82dee4c1ef06e5860b729cd087c5139d0831b997","abstract_canon_sha256":"4daddbf816c599ec59186284a19a685a308ed6752f0438e5d980eb9dfdaa86f8"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T06:15:44.429780Z","signature_b64":"xWoUw7Cmf5d+H76tnX91bpbtVjphUUgH3aCG2yVdaUpb37KOGvC8Q+WDCMSWk86/THbsQ+h6k9I32C3AkLepBQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"47512e542c702609cdda835cb7d546d3c4af46b17186ad0c9aa74c8ad4d86374","last_reissued_at":"2026-07-05T06:15:44.429267Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T06:15:44.429267Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Transformers learn in-context by gradient descent","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.LG","authors_text":"Alexander Mordvintsev, Andrey Zhmoginov, Ettore Randazzo, Eyvind Niklasson, Jo\\~ao Sacramento, Johannes von Oswald, Max Vladymyrov","submitted_at":"2022-12-15T09:21:21Z","abstract_excerpt":"At present, the mechanisms of in-context learning in Transformers are not well understood and remain mostly an intuition. In this paper, we suggest that training Transformers on auto-regressive objectives is closely related to gradient-based meta-learning formulations. We start by providing a simple weight construction that shows the equivalence of data transformations induced by 1) a single linear self-attention layer and by 2) gradient-descent (GD) on a regression loss. Motivated by that construction, we show empirically that when training self-attention-only Transformers on simple regressio"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2212.07677","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/2212.07677/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":"2212.07677","created_at":"2026-07-05T06:15:44.429340+00:00"},{"alias_kind":"arxiv_version","alias_value":"2212.07677v2","created_at":"2026-07-05T06:15:44.429340+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2212.07677","created_at":"2026-07-05T06:15:44.429340+00:00"},{"alias_kind":"pith_short_12","alias_value":"I5IS4VBMOATA","created_at":"2026-07-05T06:15:44.429340+00:00"},{"alias_kind":"pith_short_16","alias_value":"I5IS4VBMOATATTO2","created_at":"2026-07-05T06:15:44.429340+00:00"},{"alias_kind":"pith_short_8","alias_value":"I5IS4VBM","created_at":"2026-07-05T06:15:44.429340+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":7,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.26749","citing_title":"Structure Before Collapse: Transient semantic geometry in next-token prediction","ref_index":127,"is_internal_anchor":false},{"citing_arxiv_id":"2606.07623","citing_title":"Finite Certificates for In-Context Determinacy and a Threshold Theory of Emergence in Language Models","ref_index":27,"is_internal_anchor":false},{"citing_arxiv_id":"2303.17491","citing_title":"Language Models can Solve Computer Tasks","ref_index":69,"is_internal_anchor":false},{"citing_arxiv_id":"2401.10774","citing_title":"Medusa: Simple LLM Inference Acceleration Framework with Multiple Decoding Heads","ref_index":171,"is_internal_anchor":false},{"citing_arxiv_id":"2605.09727","citing_title":"One for All: A Non-Linear Transformer can Enable Cross-Domain Generalization for In-Context Reinforcement Learning","ref_index":18,"is_internal_anchor":false},{"citing_arxiv_id":"2604.23371","citing_title":"When Context Sticks: Studying Interference in In-Context Learning","ref_index":19,"is_internal_anchor":false},{"citing_arxiv_id":"2605.06322","citing_title":"SMolLM: Small Language Models Learn Small Molecular Grammar","ref_index":98,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/I5IS4VBMOATATTO2QNOLPVKG2P","json":"https://pith.science/pith/I5IS4VBMOATATTO2QNOLPVKG2P.json","graph_json":"https://pith.science/api/pith-number/I5IS4VBMOATATTO2QNOLPVKG2P/graph.json","events_json":"https://pith.science/api/pith-number/I5IS4VBMOATATTO2QNOLPVKG2P/events.json","paper":"https://pith.science/paper/I5IS4VBM"},"agent_actions":{"view_html":"https://pith.science/pith/I5IS4VBMOATATTO2QNOLPVKG2P","download_json":"https://pith.science/pith/I5IS4VBMOATATTO2QNOLPVKG2P.json","view_paper":"https://pith.science/paper/I5IS4VBM","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2212.07677&json=true","fetch_graph":"https://pith.science/api/pith-number/I5IS4VBMOATATTO2QNOLPVKG2P/graph.json","fetch_events":"https://pith.science/api/pith-number/I5IS4VBMOATATTO2QNOLPVKG2P/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/I5IS4VBMOATATTO2QNOLPVKG2P/action/timestamp_anchor","attest_storage":"https://pith.science/pith/I5IS4VBMOATATTO2QNOLPVKG2P/action/storage_attestation","attest_author":"https://pith.science/pith/I5IS4VBMOATATTO2QNOLPVKG2P/action/author_attestation","sign_citation":"https://pith.science/pith/I5IS4VBMOATATTO2QNOLPVKG2P/action/citation_signature","submit_replication":"https://pith.science/pith/I5IS4VBMOATATTO2QNOLPVKG2P/action/replication_record"}},"created_at":"2026-07-05T06:15:44.429340+00:00","updated_at":"2026-07-05T06:15:44.429340+00:00"}