{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:37763PWRJT5FWWNBCASQQ2G7KF","short_pith_number":"pith:37763PWR","schema_version":"1.0","canonical_sha256":"dfffedbed14cfa5b59a110250868df516f4b4d26ad19b63d7c1de8473e9dcff3","source":{"kind":"arxiv","id":"2306.09896","version":5},"attestation_state":"computed","paper":{"title":"Is Self-Repair a Silver Bullet for Code Generation?","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.PL","cs.SE"],"primary_cat":"cs.CL","authors_text":"Armando Solar-Lezama, Chenglong Wang, Jeevana Priya Inala, Jianfeng Gao, Theo X. Olausson","submitted_at":"2023-06-16T15:13:17Z","abstract_excerpt":"Large language models have shown remarkable aptitude in code generation, but still struggle to perform complex tasks. Self-repair -- in which the model debugs and repairs its own code -- has recently become a popular way to boost performance in these settings. However, despite its increasing popularity, existing studies of self-repair have been limited in scope; in many settings, its efficacy thus remains poorly understood. In this paper, we analyze Code Llama, GPT-3.5 and GPT-4's ability to perform self-repair on problems taken from HumanEval and APPS. We find that when the cost of carrying o"},"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":"2306.09896","kind":"arxiv","version":5},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2023-06-16T15:13:17Z","cross_cats_sorted":["cs.AI","cs.PL","cs.SE"],"title_canon_sha256":"8aa8e4fd2fb8010c484eb2ada3fbeff07ee5c21602bba64478ccb16ea2e9e196","abstract_canon_sha256":"971dd933114d6f0cb12013e3760a1441e7a2468a748fa35ff1145ae4d5b2973a"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T07:40:40.668173Z","signature_b64":"l13gRcjkbSmS0rrYGN6YdnF0ektHUNIhq8d1uhyh42nQXY4hWluM5U5I7lGUn7s/Gu+8rBXRfMn74FWp2poNBA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"dfffedbed14cfa5b59a110250868df516f4b4d26ad19b63d7c1de8473e9dcff3","last_reissued_at":"2026-07-05T07:40:40.667454Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T07:40:40.667454Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Is Self-Repair a Silver Bullet for Code Generation?","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.PL","cs.SE"],"primary_cat":"cs.CL","authors_text":"Armando Solar-Lezama, Chenglong Wang, Jeevana Priya Inala, Jianfeng Gao, Theo X. Olausson","submitted_at":"2023-06-16T15:13:17Z","abstract_excerpt":"Large language models have shown remarkable aptitude in code generation, but still struggle to perform complex tasks. Self-repair -- in which the model debugs and repairs its own code -- has recently become a popular way to boost performance in these settings. However, despite its increasing popularity, existing studies of self-repair have been limited in scope; in many settings, its efficacy thus remains poorly understood. In this paper, we analyze Code Llama, GPT-3.5 and GPT-4's ability to perform self-repair on problems taken from HumanEval and APPS. We find that when the cost of carrying o"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2306.09896","kind":"arxiv","version":5},"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/2306.09896/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":"2306.09896","created_at":"2026-07-05T07:40:40.667544+00:00"},{"alias_kind":"arxiv_version","alias_value":"2306.09896v5","created_at":"2026-07-05T07:40:40.667544+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2306.09896","created_at":"2026-07-05T07:40:40.667544+00:00"},{"alias_kind":"pith_short_12","alias_value":"37763PWRJT5F","created_at":"2026-07-05T07:40:40.667544+00:00"},{"alias_kind":"pith_short_16","alias_value":"37763PWRJT5FWWNB","created_at":"2026-07-05T07:40:40.667544+00:00"},{"alias_kind":"pith_short_8","alias_value":"37763PWR","created_at":"2026-07-05T07:40:40.667544+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":19,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.19317","citing_title":"Explaining Attention with Program Synthesis","ref_index":25,"is_internal_anchor":false},{"citing_arxiv_id":"2605.08455","citing_title":"CUDABeaver: Benchmarking LLM-Based Automated CUDA Debugging","ref_index":21,"is_internal_anchor":false},{"citing_arxiv_id":"2605.08717","citing_title":"Debugging the Debuggers: Failure-Anchored Structured Recovery for Software Engineering Agents","ref_index":27,"is_internal_anchor":false},{"citing_arxiv_id":"2606.19317","citing_title":"Explaining Attention with Program Synthesis","ref_index":25,"is_internal_anchor":false},{"citing_arxiv_id":"2605.26807","citing_title":"HTMLCure: Turning Browser Experience into State Guided Repair for Interactive HTML","ref_index":12,"is_internal_anchor":false},{"citing_arxiv_id":"2605.30052","citing_title":"REPOT: Recoverable Program-of-Thought via Checkpoint Repair","ref_index":13,"is_internal_anchor":false},{"citing_arxiv_id":"2606.05976","citing_title":"The Self-Correction Illusion: Role Relabeling Gates Explicit Error Flagging in Large Language Models","ref_index":23,"is_internal_anchor":false},{"citing_arxiv_id":"2601.08517","citing_title":"Closed-Loop LLM Discovery of Non-Standard Channel Priors in Vision Models","ref_index":16,"is_internal_anchor":false},{"citing_arxiv_id":"2309.03409","citing_title":"Large Language Models as Optimizers","ref_index":26,"is_internal_anchor":false},{"citing_arxiv_id":"2605.08717","citing_title":"Debugging the Debuggers: Failure-Anchored Structured Recovery for Software Engineering Agents","ref_index":27,"is_internal_anchor":false},{"citing_arxiv_id":"2310.01798","citing_title":"Large Language Models Cannot Self-Correct Reasoning Yet","ref_index":13,"is_internal_anchor":false},{"citing_arxiv_id":"2605.08455","citing_title":"CUDABeaver: Benchmarking LLM-Based Automated CUDA Debugging","ref_index":21,"is_internal_anchor":false},{"citing_arxiv_id":"2604.23940","citing_title":"Constraint-Guided Multi-Agent Decompilation for Executable Binary Recovery","ref_index":32,"is_internal_anchor":false},{"citing_arxiv_id":"2604.10508","citing_title":"How Many Tries Does It Take? Iterative Self-Repair in LLM Code Generation Across Model Scales and Benchmarks","ref_index":5,"is_internal_anchor":false},{"citing_arxiv_id":"2604.04580","citing_title":"Beyond Fixed Tests: Repository-Level Issue Resolution as Coevolution of Code and Behavioral Constraints","ref_index":39,"is_internal_anchor":false},{"citing_arxiv_id":"2403.07974","citing_title":"LiveCodeBench: Holistic and Contamination Free Evaluation of Large Language Models for Code","ref_index":289,"is_internal_anchor":false},{"citing_arxiv_id":"2604.16198","citing_title":"Bridging the Gap between User Intent and LLM: A Requirement Alignment Approach for Code Generation","ref_index":36,"is_internal_anchor":false},{"citing_arxiv_id":"2604.17351","citing_title":"SOCIA-EVO: Automated Simulator Construction via Dual-Anchored Bi-Level Optimization","ref_index":96,"is_internal_anchor":false},{"citing_arxiv_id":"2604.21715","citing_title":"Automated LTL Specification Generation from Industrial Aerospace Requirements","ref_index":29,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/37763PWRJT5FWWNBCASQQ2G7KF","json":"https://pith.science/pith/37763PWRJT5FWWNBCASQQ2G7KF.json","graph_json":"https://pith.science/api/pith-number/37763PWRJT5FWWNBCASQQ2G7KF/graph.json","events_json":"https://pith.science/api/pith-number/37763PWRJT5FWWNBCASQQ2G7KF/events.json","paper":"https://pith.science/paper/37763PWR"},"agent_actions":{"view_html":"https://pith.science/pith/37763PWRJT5FWWNBCASQQ2G7KF","download_json":"https://pith.science/pith/37763PWRJT5FWWNBCASQQ2G7KF.json","view_paper":"https://pith.science/paper/37763PWR","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2306.09896&json=true","fetch_graph":"https://pith.science/api/pith-number/37763PWRJT5FWWNBCASQQ2G7KF/graph.json","fetch_events":"https://pith.science/api/pith-number/37763PWRJT5FWWNBCASQQ2G7KF/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/37763PWRJT5FWWNBCASQQ2G7KF/action/timestamp_anchor","attest_storage":"https://pith.science/pith/37763PWRJT5FWWNBCASQQ2G7KF/action/storage_attestation","attest_author":"https://pith.science/pith/37763PWRJT5FWWNBCASQQ2G7KF/action/author_attestation","sign_citation":"https://pith.science/pith/37763PWRJT5FWWNBCASQQ2G7KF/action/citation_signature","submit_replication":"https://pith.science/pith/37763PWRJT5FWWNBCASQQ2G7KF/action/replication_record"}},"created_at":"2026-07-05T07:40:40.667544+00:00","updated_at":"2026-07-05T07:40:40.667544+00:00"}