{"paper":{"title":"Fix Initial Programs and Iteratively Refine Repair Instructions Toward Non-Elimination Multi-Turn Program Correction","license":"http://creativecommons.org/licenses/by/4.0/","headline":"Fixing initial codes and iteratively refining textual directions achieves comparable performance to state-of-the-art code correction methods and permits a formal safety proof.","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Issei Sato, Yuto Tanaka","submitted_at":"2026-04-27T03:07:34Z","abstract_excerpt":"Recent work on large language models (LLMs) has emphasized the importance of scaling inference compute. From this perspective, the state-of-the-art method Scattered Forest Search (SFS) has been proposed, employing Monte Carlo Tree Search with carefully crafted initial seeds and textual optimization for multi-turn program correction. However, its complexity makes it unclear what factors contribute to improvements in inference performance. To address this problem, we analyze SFS and propose a simpler method, \\textsc{Iterative Refinement of Repair Instructions} (IRRI), which fixes initial program"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"Experiments on several code generation benchmarks suggest that IRTD achieves inference performance comparable to state-of-the-art methods. Because of the simplicity of IRTD, we theoretically establish the safety of IRTD using Oracle-Guided Inductive Synthesis (OGIS).","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That the theoretical safety proof via Oracle-Guided Inductive Synthesis applies directly to the practical IRTD implementation and that benchmark results generalize without detailed statistical controls or error analysis.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"IRTD fixes initial codes and iteratively refines textual directions to achieve safe multi-turn code correction with performance comparable to the more complex Scattered Forest Search method.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Fixing initial codes and iteratively refining textual directions achieves comparable performance to state-of-the-art code correction methods and permits a formal safety proof.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"3735d478e983e498c896a6e990915b33afb289bfd3e373174b325fcc73793ff6"},"source":{"id":"2604.23989","kind":"arxiv","version":2},"verdict":{"id":"feeed88d-b22f-434d-9c4e-501a2edeb61a","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-08T04:25:09.839754Z","strongest_claim":"Experiments on several code generation benchmarks suggest that IRTD achieves inference performance comparable to state-of-the-art methods. Because of the simplicity of IRTD, we theoretically establish the safety of IRTD using Oracle-Guided Inductive Synthesis (OGIS).","one_line_summary":"IRTD fixes initial codes and iteratively refines textual directions to achieve safe multi-turn code correction with performance comparable to the more complex Scattered Forest Search method.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That the theoretical safety proof via Oracle-Guided Inductive Synthesis applies directly to the practical IRTD implementation and that benchmark results generalize without detailed statistical controls or error analysis.","pith_extraction_headline":"Fixing initial codes and iteratively refining textual directions achieves comparable performance to state-of-the-art code correction methods and permits a formal safety proof."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2604.23989/integrity.json","findings":[],"available":true,"detectors_run":[{"name":"ai_meta_artifact","ran_at":"2026-05-21T07:41:10.237904Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_compliance","ran_at":"2026-05-19T22:32:58.359341Z","status":"completed","version":"1.0.0","findings_count":0}],"snapshot_sha256":"6ea6f06892288e887af5e3d0c40818f3c7b698e2a2adb04304aa6cbdf0a62a0b"},"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"}