{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:IB56YPAMHGFDCGTZDOTL5E7HGH","short_pith_number":"pith:IB56YPAM","schema_version":"1.0","canonical_sha256":"407bec3c0c398a311a791ba6be93e731d45a6aa9c4d136c14229b0cb519330d2","source":{"kind":"arxiv","id":"2409.04556","version":2},"attestation_state":"computed","paper":{"title":"How Does Code Pretraining Affect Language Model Task Performance?","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Jackson Petty, Sjoerd van Steenkiste, Tal Linzen","submitted_at":"2024-09-06T18:33:38Z","abstract_excerpt":"Large language models are increasingly trained on corpora containing both natural language and non-linguistic data like source code. Aside from aiding programming-related tasks, anecdotal evidence suggests that including code in pretraining corpora may improve performance on other, unrelated tasks, yet to date no work has been able to establish a causal connection by controlling between language and code data. Here we do just this. We pretrain language models on datasets which interleave natural language and code in two different settings: additive, in which the total volume of data seen durin"},"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":"2409.04556","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2024-09-06T18:33:38Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"15accb5ba664c700c20d59bd846ec9a0551cf486295b91c0c1facba90d05205c","abstract_canon_sha256":"fd2d9f77184d75b6357cc765499343a54a7f2e0e76f219d6c93dcac338d75b5c"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:19:28.018584Z","signature_b64":"OHuJoxw7FB+zlxyMWzJOITR6GgsdpuZfjxBvw3FiDbaRTaUomVO0LNwEBGwL6Ai+CEfrxRD/V6DwVFXGJdzHDw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"407bec3c0c398a311a791ba6be93e731d45a6aa9c4d136c14229b0cb519330d2","last_reissued_at":"2026-07-05T10:19:28.018104Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:19:28.018104Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"How Does Code Pretraining Affect Language Model Task Performance?","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Jackson Petty, Sjoerd van Steenkiste, Tal Linzen","submitted_at":"2024-09-06T18:33:38Z","abstract_excerpt":"Large language models are increasingly trained on corpora containing both natural language and non-linguistic data like source code. Aside from aiding programming-related tasks, anecdotal evidence suggests that including code in pretraining corpora may improve performance on other, unrelated tasks, yet to date no work has been able to establish a causal connection by controlling between language and code data. Here we do just this. We pretrain language models on datasets which interleave natural language and code in two different settings: additive, in which the total volume of data seen durin"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2409.04556","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/2409.04556/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":"2409.04556","created_at":"2026-07-05T10:19:28.018158+00:00"},{"alias_kind":"arxiv_version","alias_value":"2409.04556v2","created_at":"2026-07-05T10:19:28.018158+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2409.04556","created_at":"2026-07-05T10:19:28.018158+00:00"},{"alias_kind":"pith_short_12","alias_value":"IB56YPAMHGFD","created_at":"2026-07-05T10:19:28.018158+00:00"},{"alias_kind":"pith_short_16","alias_value":"IB56YPAMHGFDCGTZ","created_at":"2026-07-05T10:19:28.018158+00:00"},{"alias_kind":"pith_short_8","alias_value":"IB56YPAM","created_at":"2026-07-05T10:19:28.018158+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.05267","citing_title":"Bridging Generation and Training: A Systematic Review of Quality Issues in LLMs for Code","ref_index":100,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/IB56YPAMHGFDCGTZDOTL5E7HGH","json":"https://pith.science/pith/IB56YPAMHGFDCGTZDOTL5E7HGH.json","graph_json":"https://pith.science/api/pith-number/IB56YPAMHGFDCGTZDOTL5E7HGH/graph.json","events_json":"https://pith.science/api/pith-number/IB56YPAMHGFDCGTZDOTL5E7HGH/events.json","paper":"https://pith.science/paper/IB56YPAM"},"agent_actions":{"view_html":"https://pith.science/pith/IB56YPAMHGFDCGTZDOTL5E7HGH","download_json":"https://pith.science/pith/IB56YPAMHGFDCGTZDOTL5E7HGH.json","view_paper":"https://pith.science/paper/IB56YPAM","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2409.04556&json=true","fetch_graph":"https://pith.science/api/pith-number/IB56YPAMHGFDCGTZDOTL5E7HGH/graph.json","fetch_events":"https://pith.science/api/pith-number/IB56YPAMHGFDCGTZDOTL5E7HGH/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/IB56YPAMHGFDCGTZDOTL5E7HGH/action/timestamp_anchor","attest_storage":"https://pith.science/pith/IB56YPAMHGFDCGTZDOTL5E7HGH/action/storage_attestation","attest_author":"https://pith.science/pith/IB56YPAMHGFDCGTZDOTL5E7HGH/action/author_attestation","sign_citation":"https://pith.science/pith/IB56YPAMHGFDCGTZDOTL5E7HGH/action/citation_signature","submit_replication":"https://pith.science/pith/IB56YPAMHGFDCGTZDOTL5E7HGH/action/replication_record"}},"created_at":"2026-07-05T10:19:28.018158+00:00","updated_at":"2026-07-05T10:19:28.018158+00:00"}