{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:LKVRPJLMTMVMKK7FCDPTBG3XYQ","short_pith_number":"pith:LKVRPJLM","schema_version":"1.0","canonical_sha256":"5aab17a56c9b2ac52be510df309b77c43f30eb14868496512da1a8cc85aa022c","source":{"kind":"arxiv","id":"2503.10061","version":3},"attestation_state":"computed","paper":{"title":"Compute Optimal Scaling of Skills: Knowledge vs Reasoning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.LG","authors_text":"Dieuwke Hupkes, Mike Lewis, Nicholas Roberts, Niladri Chatterji, Sharan Narang","submitted_at":"2025-03-13T05:21:22Z","abstract_excerpt":"Scaling laws are a critical component of the LLM development pipeline, most famously as a way to forecast training decisions such as 'compute-optimally' trading-off parameter count and dataset size, alongside a more recent growing list of other crucial decisions. In this work, we ask whether compute-optimal scaling behaviour can be skill-dependent. In particular, we examine knowledge and reasoning-based skills such as knowledge-based QA and code generation, and we answer this question in the affirmative: scaling laws are skill-dependent. Next, to understand whether skill-dependent scaling is a"},"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":"2503.10061","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-03-13T05:21:22Z","cross_cats_sorted":["cs.AI","cs.CL"],"title_canon_sha256":"4951b714c288f0d6c25f05c535dee18dad57b4f7bb99c0322eb6a34aa3317911","abstract_canon_sha256":"daa030477aef8264f6ba4cb5bf08ba245aa158573e1daf5dd68bb41c2e8f1433"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:21:15.308601Z","signature_b64":"R9EPqsS3O7vDqCT4Xze6glP+S143TD69Ijg0VEI5qdw0iSinJJ7HNsEYwsq9zNI/aj2N7dtoiKlqxRzw74YkAg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"5aab17a56c9b2ac52be510df309b77c43f30eb14868496512da1a8cc85aa022c","last_reissued_at":"2026-07-05T11:21:15.308126Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:21:15.308126Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Compute Optimal Scaling of Skills: Knowledge vs Reasoning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.LG","authors_text":"Dieuwke Hupkes, Mike Lewis, Nicholas Roberts, Niladri Chatterji, Sharan Narang","submitted_at":"2025-03-13T05:21:22Z","abstract_excerpt":"Scaling laws are a critical component of the LLM development pipeline, most famously as a way to forecast training decisions such as 'compute-optimally' trading-off parameter count and dataset size, alongside a more recent growing list of other crucial decisions. In this work, we ask whether compute-optimal scaling behaviour can be skill-dependent. In particular, we examine knowledge and reasoning-based skills such as knowledge-based QA and code generation, and we answer this question in the affirmative: scaling laws are skill-dependent. Next, to understand whether skill-dependent scaling is a"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2503.10061","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/2503.10061/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":"2503.10061","created_at":"2026-07-05T11:21:15.308184+00:00"},{"alias_kind":"arxiv_version","alias_value":"2503.10061v3","created_at":"2026-07-05T11:21:15.308184+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2503.10061","created_at":"2026-07-05T11:21:15.308184+00:00"},{"alias_kind":"pith_short_12","alias_value":"LKVRPJLMTMVM","created_at":"2026-07-05T11:21:15.308184+00:00"},{"alias_kind":"pith_short_16","alias_value":"LKVRPJLMTMVMKK7F","created_at":"2026-07-05T11:21:15.308184+00:00"},{"alias_kind":"pith_short_8","alias_value":"LKVRPJLM","created_at":"2026-07-05T11:21:15.308184+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2509.22343","citing_title":"Transformers Can Learn Connectivity in Some Graphs but Not Others","ref_index":16,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/LKVRPJLMTMVMKK7FCDPTBG3XYQ","json":"https://pith.science/pith/LKVRPJLMTMVMKK7FCDPTBG3XYQ.json","graph_json":"https://pith.science/api/pith-number/LKVRPJLMTMVMKK7FCDPTBG3XYQ/graph.json","events_json":"https://pith.science/api/pith-number/LKVRPJLMTMVMKK7FCDPTBG3XYQ/events.json","paper":"https://pith.science/paper/LKVRPJLM"},"agent_actions":{"view_html":"https://pith.science/pith/LKVRPJLMTMVMKK7FCDPTBG3XYQ","download_json":"https://pith.science/pith/LKVRPJLMTMVMKK7FCDPTBG3XYQ.json","view_paper":"https://pith.science/paper/LKVRPJLM","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2503.10061&json=true","fetch_graph":"https://pith.science/api/pith-number/LKVRPJLMTMVMKK7FCDPTBG3XYQ/graph.json","fetch_events":"https://pith.science/api/pith-number/LKVRPJLMTMVMKK7FCDPTBG3XYQ/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/LKVRPJLMTMVMKK7FCDPTBG3XYQ/action/timestamp_anchor","attest_storage":"https://pith.science/pith/LKVRPJLMTMVMKK7FCDPTBG3XYQ/action/storage_attestation","attest_author":"https://pith.science/pith/LKVRPJLMTMVMKK7FCDPTBG3XYQ/action/author_attestation","sign_citation":"https://pith.science/pith/LKVRPJLMTMVMKK7FCDPTBG3XYQ/action/citation_signature","submit_replication":"https://pith.science/pith/LKVRPJLMTMVMKK7FCDPTBG3XYQ/action/replication_record"}},"created_at":"2026-07-05T11:21:15.308184+00:00","updated_at":"2026-07-05T11:21:15.308184+00:00"}