The integration of Large Language Models (LLMs) into software engineering practices continues to evolve, with a particular focus on enhancing productivity and reducing cognitive strain. Recent advancements highlight the need for more nuanced approaches to leveraging LLMs, particularly in tasks such as code review, repository mining, and collaborative problem solving. Innovative methods are being developed to fine-tune and personalize LLM outputs, ensuring they meet specific developer needs and improve the accuracy of tasks like code readability evaluation. Additionally, there is a growing emphasis on detecting and mitigating inconsistencies in API documentation through advanced symbolic execution and LLM-assisted analysis. Noteworthy is the exploration of generative AI in root cause analysis for legacy systems, offering a proactive approach to incident resolution. However, challenges remain in ensuring the reliability and cost-effectiveness of LLM applications, with ongoing research needed to address issues such as hallucinations and model biases. Overall, the field is moving towards more personalized, accurate, and proactive uses of LLMs, with a strong focus on improving human-AI interaction and system reliability.
Enhancing Software Engineering with Personalized and Proactive LLM Applications
Sources
LLMs are Imperfect, Then What? An Empirical Study on LLM Failures in Software Engineering
Experiences from Using LLMs for Repository Mining Studies in Empirical Software Engineering
Prompting and Fine-tuning Large Language Models for Automated Code Review Comment Generation
Scaling up the Evaluation of Collaborative Problem Solving: Promises and Challenges of Coding Chat Data with ChatGPT
Personalization of Code Readability Evaluation Based on LLM Using Collaborative Filtering
Detecting Multi-Parameter Constraint Inconsistencies in Python Data Science Libraries
Deep Learning-based Code Reviews: A Paradigm Shift or a Double-Edged Sword?
LLM4DS: Evaluating Large Language Models for Data Science Code Generation
Breaking the Cycle of Recurring Failures: Applying Generative AI to Root Cause Analysis in Legacy Banking Systems
Utilizing Large Language Models to Synthesize Product Desirability Datasets
SpecTool: A Benchmark for Characterizing Errors in Tool-Use LLMs
Evidence is All We Need: Do Self-Admitted Technical Debts Impact Method-Level Maintenance?
FunctionChat-Bench: Comprehensive Evaluation of Language Models' Generative Capabilities in Korean Tool-use Dialogs