Fix Initial Codes and Iteratively Refine Textual Directions Toward Safe Multi-Turn Code Correction

2026-04-26

AI Summary

This work addresses the unclear performance gains and structural complexity of existing complex reasoning methods—such as Self-Feedback Search (SFS)—in multi-round code repair, which often compromise safety guarantees. The authors propose a concise and efficient iterative text-based refinement mechanism that fixes the initial generated code and iteratively improves it solely through high-quality textual feedback. Grounded in Oracle-Guided Inductive Synthesis (OGIS) theory, this approach provides rigorous theoretical safety assurances while avoiding intricate search architectures. Empirical evaluations demonstrate that the method achieves performance on par with state-of-the-art techniques across multiple code generation benchmarks, offering both interpretability and formal correctness guarantees.

Technology Category

• Natural Language Processing: Code Generation / Program Synthesis from Natural Language
• Search and Optimization: Sampling/Simulation-based Search
• Constraint Satisfaction and Optimization: Satisfiability Modulo Theories

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Economics, Online Markets and Human Computation: LLM based quality controls for crowd work

Abstract

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 code 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, Iterative Refinement of Textual Directions (IRTD), which fixes initial codes and iteratively refines textual directions. Because of the simplicity of IRTD, we theoretically establish the safety of IRTD using Oracle-Guided Inductive Synthesis (OGIS). Experiments on several code generation benchmarks suggest that IRTD achieves inference performance comparable to state-of-the-art methods. These results indicate that, even without complex search structures, refining initial codes with high-quality textual directions alone can effectively improve inference performance.

Problem

• multi-turn code correction
• inference performance
• large language models
• code generation
• textual directions

Innovation

• Iterative Refinement
• Textual Directions
• Code Correction
• Oracle-Guided Inductive Synthesis
• Safe Inference