{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:OHLYRMV64NY7IWUMJOPP3ZVGES","short_pith_number":"pith:OHLYRMV6","schema_version":"1.0","canonical_sha256":"71d788b2bee371f45a8c4b9efde6a62485a4b18c0efa461894642142172b0798","source":{"kind":"arxiv","id":"2603.23994","version":2},"attestation_state":"computed","paper":{"title":"Understanding the Challenges in Iterative Generative Optimization with LLMs","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Abhinav Akkiraju, Adith Swaminathan, Allen Nie, Anish Chaudhuri, Ching-An Cheng, Max Piasevoli, Prerit Choudhary, Rasool Fakoor, Ryan Rong, Shannon Xiao, Xavier Daull, YuCheng Yuan, Zhiyi Kuang","submitted_at":"2026-03-25T06:49:24Z","abstract_excerpt":"Generative optimization uses large language models (LLMs) to iteratively improve artifacts (such as code, workflows or prompts) using execution feedback. It is a promising approach to building self-improving agents, yet in practice remains brittle: despite active research, only 9% of surveyed agents used any automated optimization. We argue that this brittleness arises because, to set up a learning loop, an engineer must make ``hidden'' design choices: What can the optimizer edit and what is the \"right\" learning evidence to provide at each update? We investigate three factors that affect most "},"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":"2603.23994","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-03-25T06:49:24Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"417bb8c841b4abce63f45b0d0f823b90d8053ff8c86ce95dc488de71f512ce46","abstract_canon_sha256":"07b4697affbf8512311a67420b8d29b835d218857b7029b858a2658d793270d8"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-27T02:06:13.855032Z","signature_b64":"jY5jsN43UaC8d3abAaSMf3pw0rUfZ7fGYaa21Ed8MfqVU3QJ3RyiBcBjFhckmiFCQmIVaHyUcmi2TuAhFuLHCw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"71d788b2bee371f45a8c4b9efde6a62485a4b18c0efa461894642142172b0798","last_reissued_at":"2026-05-27T02:06:13.854235Z","signature_status":"signed_v1","first_computed_at":"2026-05-27T02:06:13.854235Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Understanding the Challenges in Iterative Generative Optimization with LLMs","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Abhinav Akkiraju, Adith Swaminathan, Allen Nie, Anish Chaudhuri, Ching-An Cheng, Max Piasevoli, Prerit Choudhary, Rasool Fakoor, Ryan Rong, Shannon Xiao, Xavier Daull, YuCheng Yuan, Zhiyi Kuang","submitted_at":"2026-03-25T06:49:24Z","abstract_excerpt":"Generative optimization uses large language models (LLMs) to iteratively improve artifacts (such as code, workflows or prompts) using execution feedback. It is a promising approach to building self-improving agents, yet in practice remains brittle: despite active research, only 9% of surveyed agents used any automated optimization. We argue that this brittleness arises because, to set up a learning loop, an engineer must make ``hidden'' design choices: What can the optimizer edit and what is the \"right\" learning evidence to provide at each update? We investigate three factors that affect most "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2603.23994","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/2603.23994/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":"2603.23994","created_at":"2026-05-27T02:06:13.854328+00:00"},{"alias_kind":"arxiv_version","alias_value":"2603.23994v2","created_at":"2026-05-27T02:06:13.854328+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2603.23994","created_at":"2026-05-27T02:06:13.854328+00:00"},{"alias_kind":"pith_short_12","alias_value":"OHLYRMV64NY7","created_at":"2026-05-27T02:06:13.854328+00:00"},{"alias_kind":"pith_short_16","alias_value":"OHLYRMV64NY7IWUM","created_at":"2026-05-27T02:06:13.854328+00:00"},{"alias_kind":"pith_short_8","alias_value":"OHLYRMV6","created_at":"2026-05-27T02:06:13.854328+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":3,"sample":[{"citing_arxiv_id":"2605.20086","citing_title":"What Do Evolutionary Coding Agents Evolve?","ref_index":68,"is_internal_anchor":true},{"citing_arxiv_id":"2604.14585","citing_title":"Prompt Optimization Is a Coin Flip: Diagnosing When It Helps in Compound AI Systems","ref_index":8,"is_internal_anchor":true},{"citing_arxiv_id":"2604.15877","citing_title":"Experience Compression Spectrum: Unifying Memory, Skills, and Rules in LLM Agents","ref_index":17,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/OHLYRMV64NY7IWUMJOPP3ZVGES","json":"https://pith.science/pith/OHLYRMV64NY7IWUMJOPP3ZVGES.json","graph_json":"https://pith.science/api/pith-number/OHLYRMV64NY7IWUMJOPP3ZVGES/graph.json","events_json":"https://pith.science/api/pith-number/OHLYRMV64NY7IWUMJOPP3ZVGES/events.json","paper":"https://pith.science/paper/OHLYRMV6"},"agent_actions":{"view_html":"https://pith.science/pith/OHLYRMV64NY7IWUMJOPP3ZVGES","download_json":"https://pith.science/pith/OHLYRMV64NY7IWUMJOPP3ZVGES.json","view_paper":"https://pith.science/paper/OHLYRMV6","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2603.23994&json=true","fetch_graph":"https://pith.science/api/pith-number/OHLYRMV64NY7IWUMJOPP3ZVGES/graph.json","fetch_events":"https://pith.science/api/pith-number/OHLYRMV64NY7IWUMJOPP3ZVGES/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/OHLYRMV64NY7IWUMJOPP3ZVGES/action/timestamp_anchor","attest_storage":"https://pith.science/pith/OHLYRMV64NY7IWUMJOPP3ZVGES/action/storage_attestation","attest_author":"https://pith.science/pith/OHLYRMV64NY7IWUMJOPP3ZVGES/action/author_attestation","sign_citation":"https://pith.science/pith/OHLYRMV64NY7IWUMJOPP3ZVGES/action/citation_signature","submit_replication":"https://pith.science/pith/OHLYRMV64NY7IWUMJOPP3ZVGES/action/replication_record"}},"created_at":"2026-05-27T02:06:13.854328+00:00","updated_at":"2026-05-27T02:06:13.854328+00:00"}