{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:MMDTBHGDOAV65BI5ZGZCDWOE4N","short_pith_number":"pith:MMDTBHGD","schema_version":"1.0","canonical_sha256":"6307309cc3702bee851dc9b221d9c4e375fbb9e007f00c94e583b24430729d4c","source":{"kind":"arxiv","id":"2410.02140","version":3},"attestation_state":"computed","paper":{"title":"A Formal Framework for Understanding Length Generalization in Transformers","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Andreas Krebs, Andy Yang, Hattie Zhou, Michael Hahn, Preetum Nakkiran, Satwik Bhattamishra, Xinting Huang, Yash Sarrof","submitted_at":"2024-10-03T01:52:01Z","abstract_excerpt":"A major challenge for transformers is generalizing to sequences longer than those observed during training. While previous works have empirically shown that transformers can either succeed or fail at length generalization depending on the task, theoretical understanding of this phenomenon remains limited. In this work, we introduce a rigorous theoretical framework to analyze length generalization in causal transformers with learnable absolute positional encodings. In particular, we characterize those functions that are identifiable in the limit from sufficiently long inputs with absolute posit"},"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":"2410.02140","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-10-03T01:52:01Z","cross_cats_sorted":[],"title_canon_sha256":"388d2dc772090d614cfa0fe2cf199c7d972118eac7ea9361068b001c21a4fdab","abstract_canon_sha256":"d0aeab13ec97c6e94bcd6404106c6ebe3bfd99d17521ae688b699d82a0c51db4"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:56:09.333937Z","signature_b64":"mA2iE7TtzxLvNvhj1beQ9dIufBookA8v3wdBYFJk0JnlxG9tKqmcX8Wczd209m2M5c6csDz/155nN1ZZcSwhCQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"6307309cc3702bee851dc9b221d9c4e375fbb9e007f00c94e583b24430729d4c","last_reissued_at":"2026-07-05T10:56:09.333440Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:56:09.333440Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"A Formal Framework for Understanding Length Generalization in Transformers","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Andreas Krebs, Andy Yang, Hattie Zhou, Michael Hahn, Preetum Nakkiran, Satwik Bhattamishra, Xinting Huang, Yash Sarrof","submitted_at":"2024-10-03T01:52:01Z","abstract_excerpt":"A major challenge for transformers is generalizing to sequences longer than those observed during training. While previous works have empirically shown that transformers can either succeed or fail at length generalization depending on the task, theoretical understanding of this phenomenon remains limited. In this work, we introduce a rigorous theoretical framework to analyze length generalization in causal transformers with learnable absolute positional encodings. In particular, we characterize those functions that are identifiable in the limit from sufficiently long inputs with absolute posit"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2410.02140","kind":"arxiv","version":3},"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/2410.02140/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":"2410.02140","created_at":"2026-07-05T10:56:09.333498+00:00"},{"alias_kind":"arxiv_version","alias_value":"2410.02140v3","created_at":"2026-07-05T10:56:09.333498+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2410.02140","created_at":"2026-07-05T10:56:09.333498+00:00"},{"alias_kind":"pith_short_12","alias_value":"MMDTBHGDOAV6","created_at":"2026-07-05T10:56:09.333498+00:00"},{"alias_kind":"pith_short_16","alias_value":"MMDTBHGDOAV65BI5","created_at":"2026-07-05T10:56:09.333498+00:00"},{"alias_kind":"pith_short_8","alias_value":"MMDTBHGD","created_at":"2026-07-05T10:56:09.333498+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.00183","citing_title":"Agentic Transformers Provably Learn to Search via Reinforcement Learning","ref_index":34,"is_internal_anchor":false},{"citing_arxiv_id":"2602.01651","citing_title":"On the Spatiotemporal Dynamics of Generalization in Neural Networks","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":32,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/MMDTBHGDOAV65BI5ZGZCDWOE4N","json":"https://pith.science/pith/MMDTBHGDOAV65BI5ZGZCDWOE4N.json","graph_json":"https://pith.science/api/pith-number/MMDTBHGDOAV65BI5ZGZCDWOE4N/graph.json","events_json":"https://pith.science/api/pith-number/MMDTBHGDOAV65BI5ZGZCDWOE4N/events.json","paper":"https://pith.science/paper/MMDTBHGD"},"agent_actions":{"view_html":"https://pith.science/pith/MMDTBHGDOAV65BI5ZGZCDWOE4N","download_json":"https://pith.science/pith/MMDTBHGDOAV65BI5ZGZCDWOE4N.json","view_paper":"https://pith.science/paper/MMDTBHGD","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2410.02140&json=true","fetch_graph":"https://pith.science/api/pith-number/MMDTBHGDOAV65BI5ZGZCDWOE4N/graph.json","fetch_events":"https://pith.science/api/pith-number/MMDTBHGDOAV65BI5ZGZCDWOE4N/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/MMDTBHGDOAV65BI5ZGZCDWOE4N/action/timestamp_anchor","attest_storage":"https://pith.science/pith/MMDTBHGDOAV65BI5ZGZCDWOE4N/action/storage_attestation","attest_author":"https://pith.science/pith/MMDTBHGDOAV65BI5ZGZCDWOE4N/action/author_attestation","sign_citation":"https://pith.science/pith/MMDTBHGDOAV65BI5ZGZCDWOE4N/action/citation_signature","submit_replication":"https://pith.science/pith/MMDTBHGDOAV65BI5ZGZCDWOE4N/action/replication_record"}},"created_at":"2026-07-05T10:56:09.333498+00:00","updated_at":"2026-07-05T10:56:09.333498+00:00"}