{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:MJIHFUJ6YBQFTGS3ZTWVQ2T2SE","short_pith_number":"pith:MJIHFUJ6","schema_version":"1.0","canonical_sha256":"625072d13ec060599a5bcced586a7a9106a8fa0bc21c808550e61361851582c9","source":{"kind":"arxiv","id":"2502.02007","version":2},"attestation_state":"computed","paper":{"title":"Reasoning Bias of Next Token Prediction Training","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Pengxiao Lin, Zhi-Qin John Xu, Zhongwang Zhang","submitted_at":"2025-02-04T04:46:41Z","abstract_excerpt":"Since the inception of Large Language Models (LLMs), the quest to efficiently train them for superior reasoning capabilities has been a pivotal challenge. The dominant training paradigm for LLMs is based on next token prediction (NTP). Alternative methodologies, called Critical Token Prediction (CTP), focused exclusively on specific critical tokens (such as the answer in Q\\&A dataset), aiming to reduce the overfitting of extraneous information and noise. Contrary to initial assumptions, our research reveals that despite NTP's exposure to noise during training, it surpasses CTP in reasoning abi"},"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":"2502.02007","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-02-04T04:46:41Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"af2d0f0b880f6042a1e5d2582d42540e116d8fd1c945df6973390f0b1059b1f7","abstract_canon_sha256":"9445659f432d3e545da67d2c4b5d084d66daca514affe49a138ee2c1e2969213"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:17:15.723326Z","signature_b64":"1JfTJzcbLHqAhNKT15oqpEPBz6/tIY1SrIp6AEa5UQMV+iss2lEZI5ou/PfmieA9pzY3Plzki8WAYywsHHjdDA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"625072d13ec060599a5bcced586a7a9106a8fa0bc21c808550e61361851582c9","last_reissued_at":"2026-07-05T10:17:15.722778Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:17:15.722778Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Reasoning Bias of Next Token Prediction Training","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Pengxiao Lin, Zhi-Qin John Xu, Zhongwang Zhang","submitted_at":"2025-02-04T04:46:41Z","abstract_excerpt":"Since the inception of Large Language Models (LLMs), the quest to efficiently train them for superior reasoning capabilities has been a pivotal challenge. The dominant training paradigm for LLMs is based on next token prediction (NTP). Alternative methodologies, called Critical Token Prediction (CTP), focused exclusively on specific critical tokens (such as the answer in Q\\&A dataset), aiming to reduce the overfitting of extraneous information and noise. Contrary to initial assumptions, our research reveals that despite NTP's exposure to noise during training, it surpasses CTP in reasoning abi"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2502.02007","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/2502.02007/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":"2502.02007","created_at":"2026-07-05T10:17:15.722838+00:00"},{"alias_kind":"arxiv_version","alias_value":"2502.02007v2","created_at":"2026-07-05T10:17:15.722838+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2502.02007","created_at":"2026-07-05T10:17:15.722838+00:00"},{"alias_kind":"pith_short_12","alias_value":"MJIHFUJ6YBQF","created_at":"2026-07-05T10:17:15.722838+00:00"},{"alias_kind":"pith_short_16","alias_value":"MJIHFUJ6YBQFTGS3","created_at":"2026-07-05T10:17:15.722838+00:00"},{"alias_kind":"pith_short_8","alias_value":"MJIHFUJ6","created_at":"2026-07-05T10:17:15.722838+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/MJIHFUJ6YBQFTGS3ZTWVQ2T2SE","json":"https://pith.science/pith/MJIHFUJ6YBQFTGS3ZTWVQ2T2SE.json","graph_json":"https://pith.science/api/pith-number/MJIHFUJ6YBQFTGS3ZTWVQ2T2SE/graph.json","events_json":"https://pith.science/api/pith-number/MJIHFUJ6YBQFTGS3ZTWVQ2T2SE/events.json","paper":"https://pith.science/paper/MJIHFUJ6"},"agent_actions":{"view_html":"https://pith.science/pith/MJIHFUJ6YBQFTGS3ZTWVQ2T2SE","download_json":"https://pith.science/pith/MJIHFUJ6YBQFTGS3ZTWVQ2T2SE.json","view_paper":"https://pith.science/paper/MJIHFUJ6","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2502.02007&json=true","fetch_graph":"https://pith.science/api/pith-number/MJIHFUJ6YBQFTGS3ZTWVQ2T2SE/graph.json","fetch_events":"https://pith.science/api/pith-number/MJIHFUJ6YBQFTGS3ZTWVQ2T2SE/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/MJIHFUJ6YBQFTGS3ZTWVQ2T2SE/action/timestamp_anchor","attest_storage":"https://pith.science/pith/MJIHFUJ6YBQFTGS3ZTWVQ2T2SE/action/storage_attestation","attest_author":"https://pith.science/pith/MJIHFUJ6YBQFTGS3ZTWVQ2T2SE/action/author_attestation","sign_citation":"https://pith.science/pith/MJIHFUJ6YBQFTGS3ZTWVQ2T2SE/action/citation_signature","submit_replication":"https://pith.science/pith/MJIHFUJ6YBQFTGS3ZTWVQ2T2SE/action/replication_record"}},"created_at":"2026-07-05T10:17:15.722838+00:00","updated_at":"2026-07-05T10:17:15.722838+00:00"}