{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2021:GTZIRURJNBX3L77VNASGFVWMCS","short_pith_number":"pith:GTZIRURJ","schema_version":"1.0","canonical_sha256":"34f288d229686fb5fff5682462d6cc1494a2728b28369bd5855ae0f298516dc0","source":{"kind":"arxiv","id":"2112.01898","version":2},"attestation_state":"computed","paper":{"title":"Linear algebra with transformers","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Fran\\c{c}ois Charton","submitted_at":"2021-12-03T13:21:57Z","abstract_excerpt":"Transformers can learn to perform numerical computations from examples only. I study nine problems of linear algebra, from basic matrix operations to eigenvalue decomposition and inversion, and introduce and discuss four encoding schemes to represent real numbers. On all problems, transformers trained on sets of random matrices achieve high accuracies (over 90%). The models are robust to noise, and can generalize out of their training distribution. In particular, models trained to predict Laplace-distributed eigenvalues generalize to different classes of matrices: Wigner matrices or matrices 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":"2112.01898","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2021-12-03T13:21:57Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"09552c4fe7cc84468088868b64bd842c2915931d0963823496d999f77dd5a54b","abstract_canon_sha256":"cddcfde783ee918ccf2a901c0fd5ef40339f0e60edb944b833a83abcb0199340"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T05:13:54.804583Z","signature_b64":"sShbbU9Hj28Pzz3nsMZwkecmZGzykwzNoX5OjYcpVXv/4BM8lpHuIGVLIW4V58e27/t1cER0cwC0O9U4DuN9Aw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"34f288d229686fb5fff5682462d6cc1494a2728b28369bd5855ae0f298516dc0","last_reissued_at":"2026-07-05T05:13:54.804041Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T05:13:54.804041Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Linear algebra with transformers","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Fran\\c{c}ois Charton","submitted_at":"2021-12-03T13:21:57Z","abstract_excerpt":"Transformers can learn to perform numerical computations from examples only. I study nine problems of linear algebra, from basic matrix operations to eigenvalue decomposition and inversion, and introduce and discuss four encoding schemes to represent real numbers. On all problems, transformers trained on sets of random matrices achieve high accuracies (over 90%). The models are robust to noise, and can generalize out of their training distribution. In particular, models trained to predict Laplace-distributed eigenvalues generalize to different classes of matrices: Wigner matrices or matrices w"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2112.01898","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/2112.01898/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":"2112.01898","created_at":"2026-07-05T05:13:54.804111+00:00"},{"alias_kind":"arxiv_version","alias_value":"2112.01898v2","created_at":"2026-07-05T05:13:54.804111+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2112.01898","created_at":"2026-07-05T05:13:54.804111+00:00"},{"alias_kind":"pith_short_12","alias_value":"GTZIRURJNBX3","created_at":"2026-07-05T05:13:54.804111+00:00"},{"alias_kind":"pith_short_16","alias_value":"GTZIRURJNBX3L77V","created_at":"2026-07-05T05:13:54.804111+00:00"},{"alias_kind":"pith_short_8","alias_value":"GTZIRURJ","created_at":"2026-07-05T05:13:54.804111+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":5,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.05106","citing_title":"Arithmetic Pedagogy for Language Models","ref_index":3,"is_internal_anchor":false},{"citing_arxiv_id":"2502.12717","citing_title":"Learning the symmetric group: large from small","ref_index":4,"is_internal_anchor":false},{"citing_arxiv_id":"2605.21160","citing_title":"Learning First Integrals via Backward-Generated Data and Guided Reinforcement Learning","ref_index":26,"is_internal_anchor":false},{"citing_arxiv_id":"2604.13082","citing_title":"The Long Delay to Arithmetic Generalization: When Learned Representations Outrun Behavior","ref_index":5,"is_internal_anchor":false},{"citing_arxiv_id":"2605.01072","citing_title":"Reconstructing conformal field theoretical compositions with Transformers","ref_index":38,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/GTZIRURJNBX3L77VNASGFVWMCS","json":"https://pith.science/pith/GTZIRURJNBX3L77VNASGFVWMCS.json","graph_json":"https://pith.science/api/pith-number/GTZIRURJNBX3L77VNASGFVWMCS/graph.json","events_json":"https://pith.science/api/pith-number/GTZIRURJNBX3L77VNASGFVWMCS/events.json","paper":"https://pith.science/paper/GTZIRURJ"},"agent_actions":{"view_html":"https://pith.science/pith/GTZIRURJNBX3L77VNASGFVWMCS","download_json":"https://pith.science/pith/GTZIRURJNBX3L77VNASGFVWMCS.json","view_paper":"https://pith.science/paper/GTZIRURJ","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2112.01898&json=true","fetch_graph":"https://pith.science/api/pith-number/GTZIRURJNBX3L77VNASGFVWMCS/graph.json","fetch_events":"https://pith.science/api/pith-number/GTZIRURJNBX3L77VNASGFVWMCS/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/GTZIRURJNBX3L77VNASGFVWMCS/action/timestamp_anchor","attest_storage":"https://pith.science/pith/GTZIRURJNBX3L77VNASGFVWMCS/action/storage_attestation","attest_author":"https://pith.science/pith/GTZIRURJNBX3L77VNASGFVWMCS/action/author_attestation","sign_citation":"https://pith.science/pith/GTZIRURJNBX3L77VNASGFVWMCS/action/citation_signature","submit_replication":"https://pith.science/pith/GTZIRURJNBX3L77VNASGFVWMCS/action/replication_record"}},"created_at":"2026-07-05T05:13:54.804111+00:00","updated_at":"2026-07-05T05:13:54.804111+00:00"}