Neural Code Generation Enhancement via Functional Overlap Reranking

arXiv:2311.03366v2 Announce Type: replace-cross
Abstract: Code Large Language Models (CodeLLMs) have marked a new era in code generation advancements. However, selecting the best solutions from all possible CodeLLM solutions remains a challenge. Previous methods frequently overlooked the intricate functional similarities and interactions between clusters, resulting in suboptimal results. In this work, we introduce \textit{SRank}, a novel reranking strategy for selecting the best solution from code generation that focuses on modeling the relationship between clusters of solutions. By quantifying the functional …

abstract arxiv challenge code code generation cs.ai cs.lg cs.se functional however interactions language language models large language large language models results solutions type via work