Maintaining software design quality is crucial for the long-term maintainability and evolution of systems. However, design issues such as poor modularity and excessive complexity often emerge as codebases grow. Developers rely on external tools, such as program analysis techniques, to identify such issues. This work leverages Large Language Models (LLMs) to develop an automated approach for analyzing and localizing design issues. Large language models have demonstrated significant performance on coding tasks, but directly leveraging them for design issue localization is challenging. Large codebases exceed typical LLM context windows, and program analysis tool outputs in non-textual modalities (e.g., graphs or interactive visualizations) are incompatible with LLMs' natural language inputs. To address these challenges, we propose LOCALIZEAGENT, a novel multi-agent framework for effective design issue localization. LOCALIZEAGENT integrates the specialized agents that (1) analyze code to identify potential code design issues, (2) transform program analysis outputs into abstraction-aware LLM-friendly natural language summaries, (3) generate context-aware prompts tailored to specific refactoring types, and (4) leverage LLMs to locate and rank the localized issues based on their relevance. Our evaluation using diverse real-world codebases demonstrates significant improvements over the baseline approaches, with LOCALIZEAGENT achieving$138 \%, 166 \%$, and 206 % relative improvements in exact-match accuracy for localizing information hiding, complexity, and modularity issues, respectively.

MoveMethod is a hallmark refactoring. Despite a plethora of research tools that recommend which methods to move and where, these recommendations do not align with how expert developers perform Movemethod. Given the extensive training of Large Language Models and their reliance upon naturalness of code, they should expertly recommend which methods are misplaced in a given class and which classes are better hosts. Our formative study of 2016 LLM recommendations revealed that LLMs give expert suggestions, yet they are unreliable: up to 80 % of the suggestions are hallucinations. We introduce the first LLM fully powered assistant for MoveMethod refactoring that automates its whole end-to-end lifecycle, from recommendation to execution. We designed novel solutions that automatically filter LLM hallucinations using static analysis from IDEs and a novel workflow that requires LLMs to be self-consistent, critique, and rank refactoring suggestions. As MoveMethod refactoring requires global, project-level reasoning, we solved the limited context size of LLMs by employing refactoring-aware retrieval augment generation (RAG). Our approach, MM-assist, synergistically combines the strengths of the LLM, IDE, static analysis, and semantic relevance. In our thorough, multi-methodology empirical evaluation, we compare MM-assist with the previous state-of-the-art approaches. MMASSIST significantly outperforms them: (i) on a benchmark widely used by other researchers, our Recall@1 and Recall@3 show a$1.7 x$improvement; (ii) on a corpus of 210 recent refactorings from Open-source software, our Recall rates improve by at least$\mathbf{2. 4 x}$. Lastly, we conducted a user study with$\mathbf{3 0}$experienced participants who used MM-ASSIST to refactor their own code for one week. They rated$\mathbf{8 2. 8 \%}$of MM-aSSIST recommendations positively. This shows that MM-ASSIST is both effective and useful.

Can we take a recurrent neural network (RNN) trained to translate between languages and augment it to support a new natural language without retraining the model from scratch? Can we fix the faulty behavior of the RNN by replacing portions associated with the faulty behavior? Recent works on decomposing a fully connected neural network (FCNN) and convolutional neural network (CNN) into modules have shown the value of engineering deep models in this manner, which is standard in traditional SE but foreign for deep learning models. However, prior works focus on the image-based multi-class classification problems and cannot be applied to RNN due to (a) different layer structures, (b) loop structures, (c) different types of input-output architectures, and (d) usage of both non-linear and logistic activation functions. In this work, we propose the first approach to decompose an RNN into modules. We study different types of RNNs, i.e., Vanilla, LSTM, and GRU. Further, we show how such RNN modules can be reused and replaced in various scenarios. We evaluate our approach against 5 canonical datasets (i.e., Math QA, Brown Corpus, Wiki-toxicity, Cline OOS, and Tatoeba) and 4 model variants for each dataset. We found that decomposing a trained model has a small cost (Accuracy: -0.6%, BLEU score: +0.10%). Also, the decomposed modules can be reused and replaced without needing to retrain.