{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:OKT3MXPAFKPA4TDHJ6Q4ILZGG4","short_pith_number":"pith:OKT3MXPA","schema_version":"1.0","canonical_sha256":"72a7b65de02a9e0e4c674fa1c42f26371f887ea1e2ba7960003dccdcb82a31c8","source":{"kind":"arxiv","id":"2306.15400","version":1},"attestation_state":"computed","paper":{"title":"Length Generalization in Arithmetic Transformers","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Carles Domingo-Enrich, Fran\\c{c}ois Charton, Samy Jelassi, St\\'ephane d'Ascoli, Yuanzhi Li, Yuhuai Wu","submitted_at":"2023-06-27T11:53:25Z","abstract_excerpt":"We examine how transformers cope with two challenges: learning basic integer arithmetic, and generalizing to longer sequences than seen during training. We find that relative position embeddings enable length generalization for simple tasks, such as addition: models trained on $5$-digit numbers can perform $15$-digit sums. However, this method fails for multiplication, and we propose train set priming: adding a few ($10$ to $50$) long sequences to the training set. We show that priming allows models trained on $5$-digit $\\times$ $3$-digit multiplications to generalize to $35\\times 3$ examples."},"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":"2306.15400","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2023-06-27T11:53:25Z","cross_cats_sorted":[],"title_canon_sha256":"a80f456e7eefa46f6fb0249673f9d8d1af240f558a7115082342944d78440bfb","abstract_canon_sha256":"2b2e218c1bf54ece24617629d34149c7ec0aeec007073f1198ef315e481be135"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T06:24:58.795434Z","signature_b64":"P9hxS8Qz0gAXlTXouC0JSslLpiHhxKGM9TjcuQWJ40qmuSJq9ukvihofaET6ZxtHZbjRqZPSc3oey3d2SSAAAg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"72a7b65de02a9e0e4c674fa1c42f26371f887ea1e2ba7960003dccdcb82a31c8","last_reissued_at":"2026-07-05T06:24:58.794944Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T06:24:58.794944Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Length Generalization in Arithmetic Transformers","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Carles Domingo-Enrich, Fran\\c{c}ois Charton, Samy Jelassi, St\\'ephane d'Ascoli, Yuanzhi Li, Yuhuai Wu","submitted_at":"2023-06-27T11:53:25Z","abstract_excerpt":"We examine how transformers cope with two challenges: learning basic integer arithmetic, and generalizing to longer sequences than seen during training. We find that relative position embeddings enable length generalization for simple tasks, such as addition: models trained on $5$-digit numbers can perform $15$-digit sums. However, this method fails for multiplication, and we propose train set priming: adding a few ($10$ to $50$) long sequences to the training set. We show that priming allows models trained on $5$-digit $\\times$ $3$-digit multiplications to generalize to $35\\times 3$ examples."},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2306.15400","kind":"arxiv","version":1},"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/2306.15400/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":"2306.15400","created_at":"2026-07-05T06:24:58.795010+00:00"},{"alias_kind":"arxiv_version","alias_value":"2306.15400v1","created_at":"2026-07-05T06:24:58.795010+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2306.15400","created_at":"2026-07-05T06:24:58.795010+00:00"},{"alias_kind":"pith_short_12","alias_value":"OKT3MXPAFKPA","created_at":"2026-07-05T06:24:58.795010+00:00"},{"alias_kind":"pith_short_16","alias_value":"OKT3MXPAFKPA4TDH","created_at":"2026-07-05T06:24:58.795010+00:00"},{"alias_kind":"pith_short_8","alias_value":"OKT3MXPA","created_at":"2026-07-05T06:24:58.795010+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":7,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.21974","citing_title":"Fine-Tuning Large Language Models for Quantum Reasoning","ref_index":52,"is_internal_anchor":false},{"citing_arxiv_id":"2605.08022","citing_title":"Globally Optimal Training of Spiking Neural Networks via Parameter Reconstruction","ref_index":19,"is_internal_anchor":false},{"citing_arxiv_id":"2604.13082","citing_title":"The Long Delay to Arithmetic Generalization: When Learned Representations Outrun Behavior","ref_index":9,"is_internal_anchor":false},{"citing_arxiv_id":"2603.29069","citing_title":"On the Mirage of Long-Range Dependency, with an Application to Integer Multiplication","ref_index":30,"is_internal_anchor":false},{"citing_arxiv_id":"2605.05176","citing_title":"Understanding In-Context Learning for Nonlinear Regression with Transformers: Attention as Featurizer","ref_index":18,"is_internal_anchor":false},{"citing_arxiv_id":"2605.08022","citing_title":"Globally Optimal Training of Spiking Neural Networks via Parameter Reconstruction","ref_index":19,"is_internal_anchor":false},{"citing_arxiv_id":"2604.15306","citing_title":"Generalization in LLM Problem Solving: The Case of the Shortest Path","ref_index":24,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/OKT3MXPAFKPA4TDHJ6Q4ILZGG4","json":"https://pith.science/pith/OKT3MXPAFKPA4TDHJ6Q4ILZGG4.json","graph_json":"https://pith.science/api/pith-number/OKT3MXPAFKPA4TDHJ6Q4ILZGG4/graph.json","events_json":"https://pith.science/api/pith-number/OKT3MXPAFKPA4TDHJ6Q4ILZGG4/events.json","paper":"https://pith.science/paper/OKT3MXPA"},"agent_actions":{"view_html":"https://pith.science/pith/OKT3MXPAFKPA4TDHJ6Q4ILZGG4","download_json":"https://pith.science/pith/OKT3MXPAFKPA4TDHJ6Q4ILZGG4.json","view_paper":"https://pith.science/paper/OKT3MXPA","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2306.15400&json=true","fetch_graph":"https://pith.science/api/pith-number/OKT3MXPAFKPA4TDHJ6Q4ILZGG4/graph.json","fetch_events":"https://pith.science/api/pith-number/OKT3MXPAFKPA4TDHJ6Q4ILZGG4/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/OKT3MXPAFKPA4TDHJ6Q4ILZGG4/action/timestamp_anchor","attest_storage":"https://pith.science/pith/OKT3MXPAFKPA4TDHJ6Q4ILZGG4/action/storage_attestation","attest_author":"https://pith.science/pith/OKT3MXPAFKPA4TDHJ6Q4ILZGG4/action/author_attestation","sign_citation":"https://pith.science/pith/OKT3MXPAFKPA4TDHJ6Q4ILZGG4/action/citation_signature","submit_replication":"https://pith.science/pith/OKT3MXPAFKPA4TDHJ6Q4ILZGG4/action/replication_record"}},"created_at":"2026-07-05T06:24:58.795010+00:00","updated_at":"2026-07-05T06:24:58.795010+00:00"}