{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:GCJRPYBHBO2HBURBLU3AAVRXEE","short_pith_number":"pith:GCJRPYBH","schema_version":"1.0","canonical_sha256":"309317e0270bb470d2215d360056372108d5fa53c72ae1d769ed186b012835cc","source":{"kind":"arxiv","id":"2406.11695","version":2},"attestation_state":"computed","paper":{"title":"Optimizing Instructions and Demonstrations for Multi-Stage Language Model Programs","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Christopher Potts, David Broman, Josh Purtell, Krista Opsahl-Ong, Matei Zaharia, Michael J Ryan, Omar Khattab","submitted_at":"2024-06-17T16:12:03Z","abstract_excerpt":"Language Model Programs, i.e. sophisticated pipelines of modular language model (LM) calls, are increasingly advancing NLP tasks, but they require crafting prompts that are jointly effective for all modules. We study prompt optimization for LM programs, i.e. how to update these prompts to maximize a downstream metric without access to module-level labels or gradients. To make this tractable, we factorize our problem into optimizing the free-form instructions and few-shot demonstrations of every module and introduce several strategies to craft task-grounded instructions and navigate credit assi"},"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":"2406.11695","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2024-06-17T16:12:03Z","cross_cats_sorted":["cs.AI","cs.LG"],"title_canon_sha256":"c64e1161d4e66a6028cf376519f0e9556c06b6bc3a35ee41770d2e42c95d81d9","abstract_canon_sha256":"2729abf1ae09622c4b0d8088207e4f2f003102319553a1bf756ca6356c4d463a"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:16:18.328243Z","signature_b64":"TknH1NkYZPVYqNsmiEvIxbvPCMNV0S2FR0JCTOfh/vdx7KDJMJIlU6k2fe4wUAJAmYNOC6ij6pbxkIrtAaXdDw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"309317e0270bb470d2215d360056372108d5fa53c72ae1d769ed186b012835cc","last_reissued_at":"2026-07-05T09:16:18.327770Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:16:18.327770Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Optimizing Instructions and Demonstrations for Multi-Stage Language Model Programs","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Christopher Potts, David Broman, Josh Purtell, Krista Opsahl-Ong, Matei Zaharia, Michael J Ryan, Omar Khattab","submitted_at":"2024-06-17T16:12:03Z","abstract_excerpt":"Language Model Programs, i.e. sophisticated pipelines of modular language model (LM) calls, are increasingly advancing NLP tasks, but they require crafting prompts that are jointly effective for all modules. We study prompt optimization for LM programs, i.e. how to update these prompts to maximize a downstream metric without access to module-level labels or gradients. To make this tractable, we factorize our problem into optimizing the free-form instructions and few-shot demonstrations of every module and introduce several strategies to craft task-grounded instructions and navigate credit assi"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2406.11695","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/2406.11695/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":"2406.11695","created_at":"2026-07-05T09:16:18.327833+00:00"},{"alias_kind":"arxiv_version","alias_value":"2406.11695v2","created_at":"2026-07-05T09:16:18.327833+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2406.11695","created_at":"2026-07-05T09:16:18.327833+00:00"},{"alias_kind":"pith_short_12","alias_value":"GCJRPYBHBO2H","created_at":"2026-07-05T09:16:18.327833+00:00"},{"alias_kind":"pith_short_16","alias_value":"GCJRPYBHBO2HBURB","created_at":"2026-07-05T09:16:18.327833+00:00"},{"alias_kind":"pith_short_8","alias_value":"GCJRPYBH","created_at":"2026-07-05T09:16:18.327833+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":12,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.18181","citing_title":"IUU+DB: Tracking Illegal, Unreported, and Unregulated Fishing, Seafood Fraud, and Labor Abuse through LLM-driven Information Extraction","ref_index":33,"is_internal_anchor":false},{"citing_arxiv_id":"2606.10457","citing_title":"Trace2Policy: From Expert Behavior Traces to Self-Evolving Decision Agents","ref_index":45,"is_internal_anchor":false},{"citing_arxiv_id":"2606.30840","citing_title":"Contrastive Reflection for Iterative Prompt Optimization","ref_index":9,"is_internal_anchor":false},{"citing_arxiv_id":"2605.27088","citing_title":"LLMs Are Already Good Tutors: Training-Free Prompt Optimization for Pedagogical Math Tutoring","ref_index":13,"is_internal_anchor":false},{"citing_arxiv_id":"2605.27671","citing_title":"Evolving and Detecting Multi-Turn Deception using Geometric Signatures","ref_index":8,"is_internal_anchor":false},{"citing_arxiv_id":"2605.15721","citing_title":"Contexting as Recommendation: Evolutionary Collaborative Filtering for Context Engineering","ref_index":19,"is_internal_anchor":false},{"citing_arxiv_id":"2605.19633","citing_title":"optimize_anything: A Universal API for Optimizing any Text Parameter","ref_index":21,"is_internal_anchor":false},{"citing_arxiv_id":"2602.21697","citing_title":"EditFlow: Benchmarking and Optimizing Code Edit Recommendation Systems via Reconstruction of Developer Flows","ref_index":46,"is_internal_anchor":false},{"citing_arxiv_id":"2605.12484","citing_title":"Learning, Fast and Slow: Towards LLMs That Adapt Continually","ref_index":40,"is_internal_anchor":false},{"citing_arxiv_id":"2605.12484","citing_title":"Learning, Fast and Slow: Towards LLMs That Adapt Continually","ref_index":40,"is_internal_anchor":false},{"citing_arxiv_id":"2604.26258","citing_title":"FlowBot: Inducing LLM Workflows with Bilevel Optimization and Textual Gradients","ref_index":4,"is_internal_anchor":false},{"citing_arxiv_id":"2604.09418","citing_title":"Automated Instruction Revision (AIR): A Structured Comparison of Task Adaptation Strategies for LLM","ref_index":7,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/GCJRPYBHBO2HBURBLU3AAVRXEE","json":"https://pith.science/pith/GCJRPYBHBO2HBURBLU3AAVRXEE.json","graph_json":"https://pith.science/api/pith-number/GCJRPYBHBO2HBURBLU3AAVRXEE/graph.json","events_json":"https://pith.science/api/pith-number/GCJRPYBHBO2HBURBLU3AAVRXEE/events.json","paper":"https://pith.science/paper/GCJRPYBH"},"agent_actions":{"view_html":"https://pith.science/pith/GCJRPYBHBO2HBURBLU3AAVRXEE","download_json":"https://pith.science/pith/GCJRPYBHBO2HBURBLU3AAVRXEE.json","view_paper":"https://pith.science/paper/GCJRPYBH","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2406.11695&json=true","fetch_graph":"https://pith.science/api/pith-number/GCJRPYBHBO2HBURBLU3AAVRXEE/graph.json","fetch_events":"https://pith.science/api/pith-number/GCJRPYBHBO2HBURBLU3AAVRXEE/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/GCJRPYBHBO2HBURBLU3AAVRXEE/action/timestamp_anchor","attest_storage":"https://pith.science/pith/GCJRPYBHBO2HBURBLU3AAVRXEE/action/storage_attestation","attest_author":"https://pith.science/pith/GCJRPYBHBO2HBURBLU3AAVRXEE/action/author_attestation","sign_citation":"https://pith.science/pith/GCJRPYBHBO2HBURBLU3AAVRXEE/action/citation_signature","submit_replication":"https://pith.science/pith/GCJRPYBHBO2HBURBLU3AAVRXEE/action/replication_record"}},"created_at":"2026-07-05T09:16:18.327833+00:00","updated_at":"2026-07-05T09:16:18.327833+00:00"}