{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:UWHYQMYTUITQBS3RTGC2MLFPRL","short_pith_number":"pith:UWHYQMYT","schema_version":"1.0","canonical_sha256":"a58f883313a22700cb719985a62caf8afecb1b508211ced73b3243da5526501c","source":{"kind":"arxiv","id":"2309.16298","version":2},"attestation_state":"computed","paper":{"title":"At Which Training Stage Does Code Data Help LLMs Reasoning?","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Changjian Wang, Shanshan Li, Yingwei Ma, Yuanliang Zhang, Yue Liu, Yue Yu, Yu Jiang","submitted_at":"2023-09-28T09:50:27Z","abstract_excerpt":"Large Language Models (LLMs) have exhibited remarkable reasoning capabilities and become the foundation of language technologies. Inspired by the great success of code data in training LLMs, we naturally wonder at which training stage introducing code data can really help LLMs reasoning. To this end, this paper systematically explores the impact of code data on LLMs at different stages. Concretely, we introduce the code data at the pre-training stage, instruction-tuning stage, and both of them, respectively. Then, the reasoning capability of LLMs is comprehensively and fairly evaluated via six"},"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":"2309.16298","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2023-09-28T09:50:27Z","cross_cats_sorted":[],"title_canon_sha256":"bbc36da54240d4a0d859c7abc9993c5c4c6c60cb72df16128f1fd421b56f617f","abstract_canon_sha256":"0b88dc3a409d779295e2988ff8c459b08a22d32376c8e43675549c9d4dabf129"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T06:55:59.921176Z","signature_b64":"XwQE++X0mp9gpeYO0tU06DdZJvO1eSuI55j3MtupV4WZJzWCgtrWcsITgUG6caMi7AZ1eo8DvlUw7A7Xe6q6BQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"a58f883313a22700cb719985a62caf8afecb1b508211ced73b3243da5526501c","last_reissued_at":"2026-07-05T06:55:59.920789Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T06:55:59.920789Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"At Which Training Stage Does Code Data Help LLMs Reasoning?","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Changjian Wang, Shanshan Li, Yingwei Ma, Yuanliang Zhang, Yue Liu, Yue Yu, Yu Jiang","submitted_at":"2023-09-28T09:50:27Z","abstract_excerpt":"Large Language Models (LLMs) have exhibited remarkable reasoning capabilities and become the foundation of language technologies. Inspired by the great success of code data in training LLMs, we naturally wonder at which training stage introducing code data can really help LLMs reasoning. To this end, this paper systematically explores the impact of code data on LLMs at different stages. Concretely, we introduce the code data at the pre-training stage, instruction-tuning stage, and both of them, respectively. Then, the reasoning capability of LLMs is comprehensively and fairly evaluated via six"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2309.16298","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/2309.16298/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":"2309.16298","created_at":"2026-07-05T06:55:59.920846+00:00"},{"alias_kind":"arxiv_version","alias_value":"2309.16298v2","created_at":"2026-07-05T06:55:59.920846+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2309.16298","created_at":"2026-07-05T06:55:59.920846+00:00"},{"alias_kind":"pith_short_12","alias_value":"UWHYQMYTUITQ","created_at":"2026-07-05T06:55:59.920846+00:00"},{"alias_kind":"pith_short_16","alias_value":"UWHYQMYTUITQBS3R","created_at":"2026-07-05T06:55:59.920846+00:00"},{"alias_kind":"pith_short_8","alias_value":"UWHYQMYT","created_at":"2026-07-05T06:55:59.920846+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":5,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.17682","citing_title":"From Trainee to Trainer: LLM-Designed Training Environment for RL with Multi-Agent Reasoning","ref_index":68,"is_internal_anchor":false},{"citing_arxiv_id":"2606.11052","citing_title":"Attention Amnesia in Hybrid LLMs: When CoT Fine-Tuning Breaks Long-Range Recall, and How to Fix It","ref_index":68,"is_internal_anchor":false},{"citing_arxiv_id":"2605.19762","citing_title":"What Really Improves Mathematical Reasoning: Structured Reasoning Signals Beyond Pure Code","ref_index":12,"is_internal_anchor":false},{"citing_arxiv_id":"2604.03071","citing_title":"Automatic Textbook Formalization","ref_index":14,"is_internal_anchor":false},{"citing_arxiv_id":"2605.05267","citing_title":"Bridging Generation and Training: A Systematic Review of Quality Issues in LLMs for Code","ref_index":84,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/UWHYQMYTUITQBS3RTGC2MLFPRL","json":"https://pith.science/pith/UWHYQMYTUITQBS3RTGC2MLFPRL.json","graph_json":"https://pith.science/api/pith-number/UWHYQMYTUITQBS3RTGC2MLFPRL/graph.json","events_json":"https://pith.science/api/pith-number/UWHYQMYTUITQBS3RTGC2MLFPRL/events.json","paper":"https://pith.science/paper/UWHYQMYT"},"agent_actions":{"view_html":"https://pith.science/pith/UWHYQMYTUITQBS3RTGC2MLFPRL","download_json":"https://pith.science/pith/UWHYQMYTUITQBS3RTGC2MLFPRL.json","view_paper":"https://pith.science/paper/UWHYQMYT","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2309.16298&json=true","fetch_graph":"https://pith.science/api/pith-number/UWHYQMYTUITQBS3RTGC2MLFPRL/graph.json","fetch_events":"https://pith.science/api/pith-number/UWHYQMYTUITQBS3RTGC2MLFPRL/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/UWHYQMYTUITQBS3RTGC2MLFPRL/action/timestamp_anchor","attest_storage":"https://pith.science/pith/UWHYQMYTUITQBS3RTGC2MLFPRL/action/storage_attestation","attest_author":"https://pith.science/pith/UWHYQMYTUITQBS3RTGC2MLFPRL/action/author_attestation","sign_citation":"https://pith.science/pith/UWHYQMYTUITQBS3RTGC2MLFPRL/action/citation_signature","submit_replication":"https://pith.science/pith/UWHYQMYTUITQBS3RTGC2MLFPRL/action/replication_record"}},"created_at":"2026-07-05T06:55:59.920846+00:00","updated_at":"2026-07-05T06:55:59.920846+00:00"}