{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:MYKVBL4QQOM3PBO5SLGFFBLJYI","short_pith_number":"pith:MYKVBL4Q","schema_version":"1.0","canonical_sha256":"661550af908399b785dd92cc528569c208c53c71f8e54fa7931b4c1f68634a8a","source":{"kind":"arxiv","id":"2501.18916","version":2},"attestation_state":"computed","paper":{"title":"LLM Program Optimization via Retrieval Augmented Search","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Alexander Shypula, Osbert Bastani, Sagnik Anupam","submitted_at":"2025-01-31T06:34:47Z","abstract_excerpt":"Recent work has demonstrated the potential of large language models (LLMs) for program optimization, a key challenge in programming languages. We propose a blackbox adaptation method called Retrieval Augmented Search (RAS) that performs beam search over candidate optimizations; at each step, it retrieves in-context examples from a given training dataset of slow-fast program pairs to guide the LLM. Critically, we find that performing contextual retrieval based on an LLM-generated natural language description significantly outperforms retrieval based on the source code. We also propose AEGIS, 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":"2501.18916","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-01-31T06:34:47Z","cross_cats_sorted":[],"title_canon_sha256":"9c8e5b5e71b6057201ad7205584fa536956df599ec0c24bfa85602334e47bd46","abstract_canon_sha256":"8c71f8c9337004ea747c4ea3f835dbfa95f8d4f559c6e54fbe123940bbadb769"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-25T00:18:10.900636Z","signature_b64":"A/ShDeqz4pXUZIfK8To8GoPjlzCXGbyVI9lstodAkDQ02za0XqcjgBq50n1VACmOLRDmqwbkAShY4JYjJ2uQCA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"661550af908399b785dd92cc528569c208c53c71f8e54fa7931b4c1f68634a8a","last_reissued_at":"2026-06-25T00:18:10.900140Z","signature_status":"signed_v1","first_computed_at":"2026-06-25T00:18:10.900140Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"LLM Program Optimization via Retrieval Augmented Search","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Alexander Shypula, Osbert Bastani, Sagnik Anupam","submitted_at":"2025-01-31T06:34:47Z","abstract_excerpt":"Recent work has demonstrated the potential of large language models (LLMs) for program optimization, a key challenge in programming languages. We propose a blackbox adaptation method called Retrieval Augmented Search (RAS) that performs beam search over candidate optimizations; at each step, it retrieves in-context examples from a given training dataset of slow-fast program pairs to guide the LLM. Critically, we find that performing contextual retrieval based on an LLM-generated natural language description significantly outperforms retrieval based on the source code. We also propose AEGIS, a "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2501.18916","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/2501.18916/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":"2501.18916","created_at":"2026-06-25T00:18:10.900195+00:00"},{"alias_kind":"arxiv_version","alias_value":"2501.18916v2","created_at":"2026-06-25T00:18:10.900195+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2501.18916","created_at":"2026-06-25T00:18:10.900195+00:00"},{"alias_kind":"pith_short_12","alias_value":"MYKVBL4QQOM3","created_at":"2026-06-25T00:18:10.900195+00:00"},{"alias_kind":"pith_short_16","alias_value":"MYKVBL4QQOM3PBO5","created_at":"2026-06-25T00:18:10.900195+00:00"},{"alias_kind":"pith_short_8","alias_value":"MYKVBL4Q","created_at":"2026-06-25T00:18:10.900195+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2604.18026","citing_title":"RASP-Tuner: Retrieval-Augmented Soft Prompts for Context-Aware Black-Box Optimization in Non-Stationary Environments","ref_index":39,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/MYKVBL4QQOM3PBO5SLGFFBLJYI","json":"https://pith.science/pith/MYKVBL4QQOM3PBO5SLGFFBLJYI.json","graph_json":"https://pith.science/api/pith-number/MYKVBL4QQOM3PBO5SLGFFBLJYI/graph.json","events_json":"https://pith.science/api/pith-number/MYKVBL4QQOM3PBO5SLGFFBLJYI/events.json","paper":"https://pith.science/paper/MYKVBL4Q"},"agent_actions":{"view_html":"https://pith.science/pith/MYKVBL4QQOM3PBO5SLGFFBLJYI","download_json":"https://pith.science/pith/MYKVBL4QQOM3PBO5SLGFFBLJYI.json","view_paper":"https://pith.science/paper/MYKVBL4Q","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2501.18916&json=true","fetch_graph":"https://pith.science/api/pith-number/MYKVBL4QQOM3PBO5SLGFFBLJYI/graph.json","fetch_events":"https://pith.science/api/pith-number/MYKVBL4QQOM3PBO5SLGFFBLJYI/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/MYKVBL4QQOM3PBO5SLGFFBLJYI/action/timestamp_anchor","attest_storage":"https://pith.science/pith/MYKVBL4QQOM3PBO5SLGFFBLJYI/action/storage_attestation","attest_author":"https://pith.science/pith/MYKVBL4QQOM3PBO5SLGFFBLJYI/action/author_attestation","sign_citation":"https://pith.science/pith/MYKVBL4QQOM3PBO5SLGFFBLJYI/action/citation_signature","submit_replication":"https://pith.science/pith/MYKVBL4QQOM3PBO5SLGFFBLJYI/action/replication_record"}},"created_at":"2026-06-25T00:18:10.900195+00:00","updated_at":"2026-06-25T00:18:10.900195+00:00"}