{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:JOVJU5JALXNBHUW4QG2ENTYSXU","short_pith_number":"pith:JOVJU5JA","schema_version":"1.0","canonical_sha256":"4baa9a75205dda13d2dc81b446cf12bd3c376a66a094e594c3fbf283ffed3732","source":{"kind":"arxiv","id":"2410.08633","version":3},"attestation_state":"computed","paper":{"title":"Transformers Provably Solve Parity Efficiently with Chain of Thought","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["stat.ML"],"primary_cat":"cs.LG","authors_text":"Juno Kim, Taiji Suzuki","submitted_at":"2024-10-11T08:55:17Z","abstract_excerpt":"This work provides the first theoretical analysis of training transformers to solve complex problems by recursively generating intermediate states, analogous to fine-tuning for chain-of-thought (CoT) reasoning. We consider training a one-layer transformer to solve the fundamental $k$-parity problem, extending the work on RNNs by Wies et al. (2023). We establish three key results: (1) any finite-precision gradient-based algorithm, without intermediate supervision, requires substantial iterations to solve parity with finite samples. (2) In contrast, when intermediate parities are incorporated in"},"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":"2410.08633","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-10-11T08:55:17Z","cross_cats_sorted":["stat.ML"],"title_canon_sha256":"448b0bd0febfb18dfc84fa556aa49fdca8806ea8007219392fc5d14d9b0cb499","abstract_canon_sha256":"cdeb8d84d67909fa4d7ca2b803cc79e098ffbf8b1ea90b19e09ba4b6ccac8c32"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:28:24.149967Z","signature_b64":"1ivLVZ54aHtPAvncmp8m1qT+/5HIQ4xZcDghCBdEcsYTM4VuX3SrLXRDwt9vH5B+tudhy/cN+IVWQ9pW0DKbAg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"4baa9a75205dda13d2dc81b446cf12bd3c376a66a094e594c3fbf283ffed3732","last_reissued_at":"2026-07-05T10:28:24.149030Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:28:24.149030Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Transformers Provably Solve Parity Efficiently with Chain of Thought","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["stat.ML"],"primary_cat":"cs.LG","authors_text":"Juno Kim, Taiji Suzuki","submitted_at":"2024-10-11T08:55:17Z","abstract_excerpt":"This work provides the first theoretical analysis of training transformers to solve complex problems by recursively generating intermediate states, analogous to fine-tuning for chain-of-thought (CoT) reasoning. We consider training a one-layer transformer to solve the fundamental $k$-parity problem, extending the work on RNNs by Wies et al. (2023). We establish three key results: (1) any finite-precision gradient-based algorithm, without intermediate supervision, requires substantial iterations to solve parity with finite samples. (2) In contrast, when intermediate parities are incorporated in"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2410.08633","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/2410.08633/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":"2410.08633","created_at":"2026-07-05T10:28:24.149151+00:00"},{"alias_kind":"arxiv_version","alias_value":"2410.08633v3","created_at":"2026-07-05T10:28:24.149151+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2410.08633","created_at":"2026-07-05T10:28:24.149151+00:00"},{"alias_kind":"pith_short_12","alias_value":"JOVJU5JALXNB","created_at":"2026-07-05T10:28:24.149151+00:00"},{"alias_kind":"pith_short_16","alias_value":"JOVJU5JALXNBHUW4","created_at":"2026-07-05T10:28:24.149151+00:00"},{"alias_kind":"pith_short_8","alias_value":"JOVJU5JA","created_at":"2026-07-05T10:28:24.149151+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":6,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.28600","citing_title":"Transformers Provably Learn to Internalize Chain-of-Thought","ref_index":24,"is_internal_anchor":false},{"citing_arxiv_id":"2606.00183","citing_title":"Agentic Transformers Provably Learn to Search via Reinforcement Learning","ref_index":31,"is_internal_anchor":false},{"citing_arxiv_id":"2606.20937","citing_title":"Learning through Internalization","ref_index":1,"is_internal_anchor":false},{"citing_arxiv_id":"2604.24737","citing_title":"Learning to Think from Multiple Thinkers","ref_index":1,"is_internal_anchor":false},{"citing_arxiv_id":"2604.22951","citing_title":"The Power of Power Law: Asymmetry Enables Compositional Reasoning","ref_index":30,"is_internal_anchor":false},{"citing_arxiv_id":"2605.06819","citing_title":"A Theory of Online Learning with Autoregressive Chain-of-Thought Reasoning","ref_index":12,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/JOVJU5JALXNBHUW4QG2ENTYSXU","json":"https://pith.science/pith/JOVJU5JALXNBHUW4QG2ENTYSXU.json","graph_json":"https://pith.science/api/pith-number/JOVJU5JALXNBHUW4QG2ENTYSXU/graph.json","events_json":"https://pith.science/api/pith-number/JOVJU5JALXNBHUW4QG2ENTYSXU/events.json","paper":"https://pith.science/paper/JOVJU5JA"},"agent_actions":{"view_html":"https://pith.science/pith/JOVJU5JALXNBHUW4QG2ENTYSXU","download_json":"https://pith.science/pith/JOVJU5JALXNBHUW4QG2ENTYSXU.json","view_paper":"https://pith.science/paper/JOVJU5JA","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2410.08633&json=true","fetch_graph":"https://pith.science/api/pith-number/JOVJU5JALXNBHUW4QG2ENTYSXU/graph.json","fetch_events":"https://pith.science/api/pith-number/JOVJU5JALXNBHUW4QG2ENTYSXU/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/JOVJU5JALXNBHUW4QG2ENTYSXU/action/timestamp_anchor","attest_storage":"https://pith.science/pith/JOVJU5JALXNBHUW4QG2ENTYSXU/action/storage_attestation","attest_author":"https://pith.science/pith/JOVJU5JALXNBHUW4QG2ENTYSXU/action/author_attestation","sign_citation":"https://pith.science/pith/JOVJU5JALXNBHUW4QG2ENTYSXU/action/citation_signature","submit_replication":"https://pith.science/pith/JOVJU5JALXNBHUW4QG2ENTYSXU/action/replication_record"}},"created_at":"2026-07-05T10:28:24.149151+00:00","updated_at":"2026-07-05T10:28:24.149151+00:00"}