{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:S6ECUKFYDHIFKJM3XNTKR7TEP5","short_pith_number":"pith:S6ECUKFY","schema_version":"1.0","canonical_sha256":"97882a28b819d055259bbb66a8fe647f42463c76581bc475ea91a299c2b7ce0f","source":{"kind":"arxiv","id":"2507.00004","version":2},"attestation_state":"computed","paper":{"title":"A Theory of Inference Compute Scaling: Reasoning through Directed Stochastic Skill Search","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CY","cs.PF"],"primary_cat":"cs.LG","authors_text":"Anuj K. Nayak, Austin R. Ellis-Mohr, Lav R. Varshney","submitted_at":"2025-06-10T14:47:48Z","abstract_excerpt":"Large language models (LLMs) demand considerable computational, energy, and financial resources during both training and deployment. While scaling laws for training have guided much of the field's recent progress, inference costs now represent a significant and growing component of the overall resource burden, particularly for reasoning-focused models. Existing characterizations of compute-optimality that consider model size, dataset size, and inference tokens in isolation or in fixed combinations risk overlooking more efficient operating points. We introduce directed stochastic skill search ("},"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":"2507.00004","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-06-10T14:47:48Z","cross_cats_sorted":["cs.AI","cs.CY","cs.PF"],"title_canon_sha256":"b17027c111447da455fd41c0cc99662baa15f882b77519e371ca20d98832f8f3","abstract_canon_sha256":"812d239716a5792ff6b92cc1c20d2bd6933fca6465bc5c02600e69092cdddda3"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:34:52.764559Z","signature_b64":"0bq1wR+lsg4jWilnQjEIAmMRwuZf50GO+Z4alnqnN+J3zkfP9mPU9kFUH22FlW6DTRURfDhflUj9VVCCyq1bAw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"97882a28b819d055259bbb66a8fe647f42463c76581bc475ea91a299c2b7ce0f","last_reissued_at":"2026-07-05T11:34:52.764049Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:34:52.764049Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"A Theory of Inference Compute Scaling: Reasoning through Directed Stochastic Skill Search","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CY","cs.PF"],"primary_cat":"cs.LG","authors_text":"Anuj K. Nayak, Austin R. Ellis-Mohr, Lav R. Varshney","submitted_at":"2025-06-10T14:47:48Z","abstract_excerpt":"Large language models (LLMs) demand considerable computational, energy, and financial resources during both training and deployment. While scaling laws for training have guided much of the field's recent progress, inference costs now represent a significant and growing component of the overall resource burden, particularly for reasoning-focused models. Existing characterizations of compute-optimality that consider model size, dataset size, and inference tokens in isolation or in fixed combinations risk overlooking more efficient operating points. We introduce directed stochastic skill search ("},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2507.00004","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/2507.00004/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":"2507.00004","created_at":"2026-07-05T11:34:52.764110+00:00"},{"alias_kind":"arxiv_version","alias_value":"2507.00004v2","created_at":"2026-07-05T11:34:52.764110+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2507.00004","created_at":"2026-07-05T11:34:52.764110+00:00"},{"alias_kind":"pith_short_12","alias_value":"S6ECUKFYDHIF","created_at":"2026-07-05T11:34:52.764110+00:00"},{"alias_kind":"pith_short_16","alias_value":"S6ECUKFYDHIFKJM3","created_at":"2026-07-05T11:34:52.764110+00:00"},{"alias_kind":"pith_short_8","alias_value":"S6ECUKFY","created_at":"2026-07-05T11:34:52.764110+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2607.00913","citing_title":"Two AI Metrics Diverged: Will it Make All the Difference?","ref_index":4,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/S6ECUKFYDHIFKJM3XNTKR7TEP5","json":"https://pith.science/pith/S6ECUKFYDHIFKJM3XNTKR7TEP5.json","graph_json":"https://pith.science/api/pith-number/S6ECUKFYDHIFKJM3XNTKR7TEP5/graph.json","events_json":"https://pith.science/api/pith-number/S6ECUKFYDHIFKJM3XNTKR7TEP5/events.json","paper":"https://pith.science/paper/S6ECUKFY"},"agent_actions":{"view_html":"https://pith.science/pith/S6ECUKFYDHIFKJM3XNTKR7TEP5","download_json":"https://pith.science/pith/S6ECUKFYDHIFKJM3XNTKR7TEP5.json","view_paper":"https://pith.science/paper/S6ECUKFY","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2507.00004&json=true","fetch_graph":"https://pith.science/api/pith-number/S6ECUKFYDHIFKJM3XNTKR7TEP5/graph.json","fetch_events":"https://pith.science/api/pith-number/S6ECUKFYDHIFKJM3XNTKR7TEP5/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/S6ECUKFYDHIFKJM3XNTKR7TEP5/action/timestamp_anchor","attest_storage":"https://pith.science/pith/S6ECUKFYDHIFKJM3XNTKR7TEP5/action/storage_attestation","attest_author":"https://pith.science/pith/S6ECUKFYDHIFKJM3XNTKR7TEP5/action/author_attestation","sign_citation":"https://pith.science/pith/S6ECUKFYDHIFKJM3XNTKR7TEP5/action/citation_signature","submit_replication":"https://pith.science/pith/S6ECUKFYDHIFKJM3XNTKR7TEP5/action/replication_record"}},"created_at":"2026-07-05T11:34:52.764110+00:00","updated_at":"2026-07-05T11:34:52.764110+00:00"}