It is suggested that repair budgets should be treated as an explicit experimental variable, as they directly affect evaluation outcomes, computational cost, runtime, and reproducibility in LLM-based software engineering research.
Abstract
Iterative repair loops have become a core design pattern in LLM-based software engineering systems. These workflows repeatedly generate, validate, and repair artifacts using feedback such as compiler errors or test failures. Despite their widespread use, the impact of repair-loop iteration limits remains poorly understood, as most prior work adopts fixed, often arbitrary, repair budgets. We study repair-loop effectiveness across multiple software engineering tasks, including code generation, test generation, and code translation. Across several representative workflows, datasets, and contemporary low-cost LLMs, we observe a consistent pattern of diminishing returns: the first three to four repair iterations account for most achievable gains, while later iterations contribute only marginal improvements. We further find that repair behavior is influenced more strongly by workflow orchestration and feedback design than by the underlying model itself. These results suggest that repair budgets should be treated as an explicit experimental variable, as they directly affect evaluation outcomes, computational cost, runtime, and reproducibility in LLM-based software engineering research.
Research on LLM-based code generation has achieved strong performance on function-level benchmarks, yet repository-level code generation in real-world projects still faces challenges from cross-file dependencies. Existing studies usually emphasize single-round or final-round outcomes, which limits the analysis of where feedback gains concentrate and when later repair begins to offset earlier progress. We propose a dynamic evaluation framework for repository-level code generation based on DevEval’s Python tasks. The framework constructs processed test sets, evaluates multi-round repair trajectories with unit-test feedback over multiple rounds, and characterizes trajectories via round-level effectiveness, feedback-induced gain, and regression stability. However, roughly three-quarters of the gains emerged within the first three rounds, and net feedback gain turned negative by Round 5. Previously repaired errors reappeared in later iterations, with reappeared error rates from 4.5% to 8.8% across models. These findings indicate that unit-test feedback is most effective as an early-stage signal, and that practical repair systems should adopt early-round budgeting with best-so-far retention rather than open-ended loops.
Fuze Kuang, Dongmei Liu, Yong-Li Wang et al.· International Conference on...· 0 citations
This paper identifies patch verbosity as a major yet overlooked concern in LLM-based APR and proposes RECAP, a lightweight, plug-and-play adapter that attaches to existing repair frameworks after generation that achieves a substantially better size-correctness tradeoff.
Wen-Qiang Luo, J. Keung, Xiaoyu Shi et al.· 0 citations
Since the advent of Large Language Models (LLMs), practitioners have increasingly leveraged them to support their software engineering tasks, including automated code repair, showing promising results. Yet, concerns regarding reproducibility and generalizability remain largely unexplored. To further evaluate these concerns and associated impacts, we partially reproduce and conduct an exploratory Java extension of the FeedbackEval benchmark [1], which evaluates how LLMs leverage different feedback types for Python code repair. First, we partially replicate the original study on 394 repair tasks using GPT-4o and Claude 3.5 Sonnet, reproducing and observing the main qualitative trends reported in the original work. Second, we conduct an exploratory Java extension by constructing 100 erroneous repair instances from 50 Java tasks and evaluating feedback effectiveness. Our results show that previous conclusions from Python may be sensitive to benchmark construction, feedback representation, and tooling ecosystem, motivating more controlled multilingual benchmarks. Specifically, while test feedback remains the strongest feedback type in our Python replication, the same ranking is not observed in our Java extension, as simple and JUnit-based test feedback do not differ significantly. We hypothesize that differences in feedback informativeness and tooling ecosystems, such as the verbosity of test frameworks, may partly explain such a difference. Finally, lighter prompts reduce cost without significant differences in repair effectiveness. Overall, our findings confirm key trends under a partially controlled replication and highlight the need for more rigorous multilingual evaluation and careful feedback design in LLMbased repair systems.
L. Lalonde, Wassim Keddache, Thomas Perron Touchette et al.· 0 citations
The findings show that, while LLMs achieve promising results, they struggle with harder problems and with programming languages that have fewer available resources for training, and they often fail due to fundamental and easily avoidable errors that underscore the unreliability of automatically generated code.
Rodrigo Pato Nogueira, Marco Vieira, João R. Campos· Empirical Software Engineeri...· 1 citation
An automated, multi-dimensional evaluation framework for C# code generation, applying it to four state-of-the-art LLMs: GPT, Gemini, Claude, and Grok is presented and a substantial gap between correctness and quality attributes is revealed.
Seyed Mohammad Mahdi Ghalandarian, Majid Bazargani, Masoumeh Taromirad· 0 citations
Due to their black-box nature, LLMs suffer from limited explain- ability and a lack of determinism. Their usage cost can also rise, particularly with repetitive tasks on large codebases. To mitigate this, we conduct a novel empirical study targeting three domain- specific languages for transformation rules, namely Comby, GritQL, and Ast-Grep. We evaluate three LLMs (GPT-5.4, GPT-oss-120B, and Llama3.1-8B) on six diverse datasets covering four software- evolution tasks: API misuse correction, program repair, API migra- tion, and language version migration. Our results provide evidence that transformation rule synthesis moves beyond proof-of-concept with strong frontier models. GPT-5.4 achieves consistently high rule applicability rates and produces transformations closest to the ground truth across most benchmarks. Smaller and open-weight GPT-oss-120B and Llama3.1-8B models remain effective for simpler, localized changes but struggle with complex migration scenarios. We also observe non-negligible generalizability through the usage of meta-variables and through a high reuse score in the first quartile of many datasets. Finally, when compared to the anti-unification algorithm, LLMs outperform it in correctness, but underperform in rule applicability. Overall, our results show great potential for LLMs to generate sound, correct, generalizable, and reusable rules.
Axel Allain, Aymeric Blot, D. Khelladi et al.· 1 citation
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
