{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:FJOWQWKYILQJESVM5XITMRFTDR","short_pith_number":"pith:FJOWQWKY","schema_version":"1.0","canonical_sha256":"2a5d68595842e0924aacedd13644b31c5244e9f09c3641256689c0d3d4fc13a5","source":{"kind":"arxiv","id":"2502.14815","version":1},"attestation_state":"computed","paper":{"title":"Optimizing Model Selection for Compound AI Systems","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CL","cs.LG","cs.MA"],"primary_cat":"cs.AI","authors_text":"Boris Hanin, Ion Stoica, James Zou, Jared Quincy Davis, Lingjiao Chen, Matei Zaharia, Peter Bailis","submitted_at":"2025-02-20T18:36:25Z","abstract_excerpt":"Compound AI systems that combine multiple LLM calls, such as self-refine and multi-agent-debate, achieve strong performance on many AI tasks. We address a core question in optimizing compound systems: for each LLM call or module in the system, how should one decide which LLM to use? We show that these LLM choices have a large effect on quality, but the search space is exponential. We propose LLMSelector, an efficient framework for model selection in compound systems, which leverages two key empirical insights: (i) end-to-end performance is often monotonic in how well each module performs, with"},"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":"2502.14815","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.AI","submitted_at":"2025-02-20T18:36:25Z","cross_cats_sorted":["cs.CL","cs.LG","cs.MA"],"title_canon_sha256":"1291c7cfff6e6c28e1696d981880c6282c9ab2ead74e1362da9bd2f6b2bb6ac2","abstract_canon_sha256":"df0fc41468a75a633935880965c1b14831f9047868dabe2c8e05b960203e3892"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:17:32.874593Z","signature_b64":"KuyOzfnUMEuwovpNPfTLyTwGN/kmj73JXb3eBapcwENqxsYtNkbqFBV/gs2NJz50Kd6Uu/+u3ERp95WgY6AECw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"2a5d68595842e0924aacedd13644b31c5244e9f09c3641256689c0d3d4fc13a5","last_reissued_at":"2026-07-05T10:17:32.874186Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:17:32.874186Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Optimizing Model Selection for Compound AI Systems","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CL","cs.LG","cs.MA"],"primary_cat":"cs.AI","authors_text":"Boris Hanin, Ion Stoica, James Zou, Jared Quincy Davis, Lingjiao Chen, Matei Zaharia, Peter Bailis","submitted_at":"2025-02-20T18:36:25Z","abstract_excerpt":"Compound AI systems that combine multiple LLM calls, such as self-refine and multi-agent-debate, achieve strong performance on many AI tasks. We address a core question in optimizing compound systems: for each LLM call or module in the system, how should one decide which LLM to use? We show that these LLM choices have a large effect on quality, but the search space is exponential. We propose LLMSelector, an efficient framework for model selection in compound systems, which leverages two key empirical insights: (i) end-to-end performance is often monotonic in how well each module performs, with"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2502.14815","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/2502.14815/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":"2502.14815","created_at":"2026-07-05T10:17:32.874243+00:00"},{"alias_kind":"arxiv_version","alias_value":"2502.14815v1","created_at":"2026-07-05T10:17:32.874243+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2502.14815","created_at":"2026-07-05T10:17:32.874243+00:00"},{"alias_kind":"pith_short_12","alias_value":"FJOWQWKYILQJ","created_at":"2026-07-05T10:17:32.874243+00:00"},{"alias_kind":"pith_short_16","alias_value":"FJOWQWKYILQJESVM","created_at":"2026-07-05T10:17:32.874243+00:00"},{"alias_kind":"pith_short_8","alias_value":"FJOWQWKY","created_at":"2026-07-05T10:17:32.874243+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":9,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2607.07989","citing_title":"Who Broke the System? Failure Localization in LLM-Based Multi-Agent Systems","ref_index":57,"is_internal_anchor":true},{"citing_arxiv_id":"2606.18502","citing_title":"Towards Scalable Customization and Deployment of Multi-Agent Systems for Enterprise Applications","ref_index":15,"is_internal_anchor":false},{"citing_arxiv_id":"2606.01185","citing_title":"\"Skill issues'': data-centric optimization of lakehouse agents","ref_index":8,"is_internal_anchor":false},{"citing_arxiv_id":"2606.28365","citing_title":"CAMI: Cost-Aware Agent-Guided Multi-Indexing for Semantic Retrieval","ref_index":5,"is_internal_anchor":false},{"citing_arxiv_id":"2606.00765","citing_title":"FALAT: Tracing Failures in LLM Agent Trajectories via Dependency-Guided Search","ref_index":38,"is_internal_anchor":false},{"citing_arxiv_id":"2606.00774","citing_title":"SCOPE: Cost-Efficient Model Selection for Compound AI Systems under Quality Constraints","ref_index":3,"is_internal_anchor":false},{"citing_arxiv_id":"2606.04883","citing_title":"Optimizing the Cost-Quality Tradeoff of Agentic Theorem Provers in Lean","ref_index":68,"is_internal_anchor":false},{"citing_arxiv_id":"2605.00914","citing_title":"The Cost of Consensus: Isolated Self-Correction Prevails Over Unguided Homogeneous Multi-Agent Debate","ref_index":3,"is_internal_anchor":false},{"citing_arxiv_id":"2605.07935","citing_title":"TraceFix: Repairing Agent Coordination Protocols with TLA+ Counterexamples","ref_index":6,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/FJOWQWKYILQJESVM5XITMRFTDR","json":"https://pith.science/pith/FJOWQWKYILQJESVM5XITMRFTDR.json","graph_json":"https://pith.science/api/pith-number/FJOWQWKYILQJESVM5XITMRFTDR/graph.json","events_json":"https://pith.science/api/pith-number/FJOWQWKYILQJESVM5XITMRFTDR/events.json","paper":"https://pith.science/paper/FJOWQWKY"},"agent_actions":{"view_html":"https://pith.science/pith/FJOWQWKYILQJESVM5XITMRFTDR","download_json":"https://pith.science/pith/FJOWQWKYILQJESVM5XITMRFTDR.json","view_paper":"https://pith.science/paper/FJOWQWKY","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2502.14815&json=true","fetch_graph":"https://pith.science/api/pith-number/FJOWQWKYILQJESVM5XITMRFTDR/graph.json","fetch_events":"https://pith.science/api/pith-number/FJOWQWKYILQJESVM5XITMRFTDR/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/FJOWQWKYILQJESVM5XITMRFTDR/action/timestamp_anchor","attest_storage":"https://pith.science/pith/FJOWQWKYILQJESVM5XITMRFTDR/action/storage_attestation","attest_author":"https://pith.science/pith/FJOWQWKYILQJESVM5XITMRFTDR/action/author_attestation","sign_citation":"https://pith.science/pith/FJOWQWKYILQJESVM5XITMRFTDR/action/citation_signature","submit_replication":"https://pith.science/pith/FJOWQWKYILQJESVM5XITMRFTDR/action/replication_record"}},"created_at":"2026-07-05T10:17:32.874243+00:00","updated_at":"2026-07-05T10:17:32.874243+00:00"}