{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:JR3OG3MIADCLVPIRSMA7GCOSTY","short_pith_number":"pith:JR3OG3MI","schema_version":"1.0","canonical_sha256":"4c76e36d8800c4babd119301f309d29e0f5b249d878a647beeef3b376913437f","source":{"kind":"arxiv","id":"2608.10605","version":1},"attestation_state":"computed","paper":{"title":"Compute-Optimal Is Not Cluster-Optimal: Systems-Aware Scaling for Sparse Mixture-of-Experts","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Sheng Zha, Soumajyoti Sarkar, Yuxin Tang","submitted_at":"2026-08-11T07:49:00Z","abstract_excerpt":"In large-scale pretraining, the algorithm, architecture, and systems decisions are conventionally made in disconnected stages. A scaling law stage selects an architecture and training recipe, optimizing loss under compute constraints, and a separate systems stage then optimizes the implementation for hardware efficiency. In this work, we develop MOSAIC, which formulates model architecture and systems co-design as an optimization problem. MOSAIC couples a predictive scaling law with a calibrated performance model that estimates Model FLOPs Utilization (MFU), communication cost, memory footprint"},"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":"2608.10605","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-08-11T07:49:00Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"2d49a7cb4a6791ebbda3cbf276795c361f70185dee8eabfb5384ca22781c8731","abstract_canon_sha256":"70742ccf09ea78f69f91f9c80a39be6b2a01075c5255559f2a7145fe49478f6a"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-08-12T01:22:51.135079Z","signature_b64":"UhXKG7GeHMnm/v5I9WB4CLjA5v3ZzaFmr1jSy0p7DajKgcJqj6Y71LgnzH0ZiwtdOphNAChUlqTUYH/FUAKNBw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"4c76e36d8800c4babd119301f309d29e0f5b249d878a647beeef3b376913437f","last_reissued_at":"2026-08-12T01:22:51.133464Z","signature_status":"signed_v1","first_computed_at":"2026-08-12T01:22:51.133464Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Compute-Optimal Is Not Cluster-Optimal: Systems-Aware Scaling for Sparse Mixture-of-Experts","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Sheng Zha, Soumajyoti Sarkar, Yuxin Tang","submitted_at":"2026-08-11T07:49:00Z","abstract_excerpt":"In large-scale pretraining, the algorithm, architecture, and systems decisions are conventionally made in disconnected stages. A scaling law stage selects an architecture and training recipe, optimizing loss under compute constraints, and a separate systems stage then optimizes the implementation for hardware efficiency. In this work, we develop MOSAIC, which formulates model architecture and systems co-design as an optimization problem. MOSAIC couples a predictive scaling law with a calibrated performance model that estimates Model FLOPs Utilization (MFU), communication cost, memory footprint"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2608.10605","kind":"arxiv","version":1},"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/2608.10605/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":"2608.10605","created_at":"2026-08-12T01:22:51.137315+00:00"},{"alias_kind":"arxiv_version","alias_value":"2608.10605v1","created_at":"2026-08-12T01:22:51.137315+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2608.10605","created_at":"2026-08-12T01:22:51.137315+00:00"},{"alias_kind":"pith_short_12","alias_value":"JR3OG3MIADCL","created_at":"2026-08-12T01:22:51.137315+00:00"},{"alias_kind":"pith_short_16","alias_value":"JR3OG3MIADCLVPIR","created_at":"2026-08-12T01:22:51.137315+00:00"},{"alias_kind":"pith_short_8","alias_value":"JR3OG3MI","created_at":"2026-08-12T01:22:51.137315+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/JR3OG3MIADCLVPIRSMA7GCOSTY","json":"https://pith.science/pith/JR3OG3MIADCLVPIRSMA7GCOSTY.json","graph_json":"https://pith.science/api/pith-number/JR3OG3MIADCLVPIRSMA7GCOSTY/graph.json","events_json":"https://pith.science/api/pith-number/JR3OG3MIADCLVPIRSMA7GCOSTY/events.json","paper":"https://pith.science/paper/JR3OG3MI"},"agent_actions":{"view_html":"https://pith.science/pith/JR3OG3MIADCLVPIRSMA7GCOSTY","download_json":"https://pith.science/pith/JR3OG3MIADCLVPIRSMA7GCOSTY.json","view_paper":"https://pith.science/paper/JR3OG3MI","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2608.10605&json=true","fetch_graph":"https://pith.science/api/pith-number/JR3OG3MIADCLVPIRSMA7GCOSTY/graph.json","fetch_events":"https://pith.science/api/pith-number/JR3OG3MIADCLVPIRSMA7GCOSTY/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/JR3OG3MIADCLVPIRSMA7GCOSTY/action/timestamp_anchor","attest_storage":"https://pith.science/pith/JR3OG3MIADCLVPIRSMA7GCOSTY/action/storage_attestation","attest_author":"https://pith.science/pith/JR3OG3MIADCLVPIRSMA7GCOSTY/action/author_attestation","sign_citation":"https://pith.science/pith/JR3OG3MIADCLVPIRSMA7GCOSTY/action/citation_signature","submit_replication":"https://pith.science/pith/JR3OG3MIADCLVPIRSMA7GCOSTY/action/replication_record"}},"created_at":"2026-08-12T01:22:51.137315+00:00","updated_at":"2026-08-12T01:22:51.137315+00:00"}