{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:OUEP6ODUYVUO64TJOAWPPD3ODL","short_pith_number":"pith:OUEP6ODU","schema_version":"1.0","canonical_sha256":"7508ff3874c568ef7269702cf78f6e1ac1bfa16a1c1812f1efda0454b9cf8b47","source":{"kind":"arxiv","id":"2605.19102","version":1},"attestation_state":"computed","paper":{"title":"Prompt Optimization for LLM Code Generation via Reinforcement Learning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.SE","authors_text":"Ali Mohammadi Esfahani, Nafiseh Kahani, Samuel A.Ajila","submitted_at":"2026-05-18T20:42:23Z","abstract_excerpt":"Large Language Models (LLMs) can generate code from natural language, but their performance is highly sensitive to prompt formulation. We propose a reinforcement-learning-based framework that models prompt refinement as a sequential decision-making problem. A Proximal Policy Optimization (PPO) agent iteratively improves prompts using a hybrid action space that combines direct generation, genetic lexical mutation and semantic rewriting, guided by shaped rewards derived from unit-test feedback. We evaluate the framework on MBPP+, HumanEval+, and APPS using CodeT5+, CodeLLaMA, and DeepSeek-Coder "},"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":"2605.19102","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.SE","submitted_at":"2026-05-18T20:42:23Z","cross_cats_sorted":[],"title_canon_sha256":"6bd51326c1b33480a69770533d3664c8b40b307c6e7e39d6949114c715b6bd1f","abstract_canon_sha256":"2cec96d18416afb994c20839512fc927d6845a67ab5c3893a6654a8232ada203"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-20T01:05:27.230066Z","signature_b64":"uARfLUS2402itBrnuAsjYnO/hQ2erywW9cJxSrk5oJxXKASgbRRhPsMNjnlUspe8ht7L1P7yZ1g7AIhByPjLDQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"7508ff3874c568ef7269702cf78f6e1ac1bfa16a1c1812f1efda0454b9cf8b47","last_reissued_at":"2026-05-20T01:05:27.229250Z","signature_status":"signed_v1","first_computed_at":"2026-05-20T01:05:27.229250Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Prompt Optimization for LLM Code Generation via Reinforcement Learning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.SE","authors_text":"Ali Mohammadi Esfahani, Nafiseh Kahani, Samuel A.Ajila","submitted_at":"2026-05-18T20:42:23Z","abstract_excerpt":"Large Language Models (LLMs) can generate code from natural language, but their performance is highly sensitive to prompt formulation. We propose a reinforcement-learning-based framework that models prompt refinement as a sequential decision-making problem. A Proximal Policy Optimization (PPO) agent iteratively improves prompts using a hybrid action space that combines direct generation, genetic lexical mutation and semantic rewriting, guided by shaped rewards derived from unit-test feedback. We evaluate the framework on MBPP+, HumanEval+, and APPS using CodeT5+, CodeLLaMA, and DeepSeek-Coder "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.19102","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/2605.19102/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":"2605.19102","created_at":"2026-05-20T01:05:27.229403+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.19102v1","created_at":"2026-05-20T01:05:27.229403+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.19102","created_at":"2026-05-20T01:05:27.229403+00:00"},{"alias_kind":"pith_short_12","alias_value":"OUEP6ODUYVUO","created_at":"2026-05-20T01:05:27.229403+00:00"},{"alias_kind":"pith_short_16","alias_value":"OUEP6ODUYVUO64TJ","created_at":"2026-05-20T01:05:27.229403+00:00"},{"alias_kind":"pith_short_8","alias_value":"OUEP6ODU","created_at":"2026-05-20T01:05:27.229403+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/OUEP6ODUYVUO64TJOAWPPD3ODL","json":"https://pith.science/pith/OUEP6ODUYVUO64TJOAWPPD3ODL.json","graph_json":"https://pith.science/api/pith-number/OUEP6ODUYVUO64TJOAWPPD3ODL/graph.json","events_json":"https://pith.science/api/pith-number/OUEP6ODUYVUO64TJOAWPPD3ODL/events.json","paper":"https://pith.science/paper/OUEP6ODU"},"agent_actions":{"view_html":"https://pith.science/pith/OUEP6ODUYVUO64TJOAWPPD3ODL","download_json":"https://pith.science/pith/OUEP6ODUYVUO64TJOAWPPD3ODL.json","view_paper":"https://pith.science/paper/OUEP6ODU","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.19102&json=true","fetch_graph":"https://pith.science/api/pith-number/OUEP6ODUYVUO64TJOAWPPD3ODL/graph.json","fetch_events":"https://pith.science/api/pith-number/OUEP6ODUYVUO64TJOAWPPD3ODL/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/OUEP6ODUYVUO64TJOAWPPD3ODL/action/timestamp_anchor","attest_storage":"https://pith.science/pith/OUEP6ODUYVUO64TJOAWPPD3ODL/action/storage_attestation","attest_author":"https://pith.science/pith/OUEP6ODUYVUO64TJOAWPPD3ODL/action/author_attestation","sign_citation":"https://pith.science/pith/OUEP6ODUYVUO64TJOAWPPD3ODL/action/citation_signature","submit_replication":"https://pith.science/pith/OUEP6ODUYVUO64TJOAWPPD3ODL/action/replication_record"}},"created_at":"2026-05-20T01:05:27.229403+00:00","updated_at":"2026-05-20T01:05:27.229403+00:00"}