Search papers, labs, and topics across Lattice.
The paper introduces FixAudit, an iterative test-and-repair framework for competitive code generation that trains a single model to act as both a Fixer (repairs code based on failing tests) and an Auditor (generates tests to expose bugs). Unlike prior work like CURE, FixAudit iteratively refines a single candidate program and allows the Auditor to examine the code, leading to more targeted test generation. Experiments on APPS, CodeContests, and xCodeEval show that a 7B model using FixAudit outperforms a 32B baseline and improves Pass@1 by 35.1%-36.8% compared to other 7B baselines.
Forget brute-force code generation: FixAudit shows iteratively debugging a single candidate with targeted tests can beat much larger models.
Large language models (LLMs) have made remarkable progress in code generation, but competitive programming remains a challenge. Recent training-based methods have improved code generation by using reinforcement learning (RL) with execution feedback. The more recent framework CURE further incorporates test generation into the training process, jointly training a Coder and a Tester within a single model. At inference time, the Coder generates many candidate programs, and the Tester generates tests from the problem description. The candidate who passes the most of the generated tests is selected as the final answer. However, CURE has two critical limitations. First, the Tester never reads any candidate code, so its tests often fail to expose implementation-specific bugs. Second, the Coder generates every candidate from scratch and never learns to fix a buggy program based on a failed test. To address these limitations, we propose FixAudit, which approaches competitive code generation from a new perspective: starting from a single initial candidate, it iteratively improves the candidate through a targeted test-and-repair debugging cycle. The framework trains one shared model with two specialized roles through four stages: the Fixer, which repairs the current candidate based on a failing test, and the Auditor, which reads the candidate code to generate new tests that expose its remaining bugs. We evaluate FixAudit on three benchmarks: APPS, CodeContests, and xCodeEval. Applied to a 7B model, the framework surpasses the average performance of the larger 32B baseline within the same model family under the zero-shot setting. Compared to strong baselines built on the same 7B base model, FixAudit improves average Pass@1 by 35.1% to 36.8% and average AvgPassRatio by 7.1% to 24.5%.
