Enhancing Software Engineering Through Closed-Loop Memory Optimization

arXiv:2606.05646v1 Announce Type: cross Abstract: Large language models (LLMs) have enabled powerful software engineering (SE) agents capable of navigating complex codebases and resolving real-world issues. However, these agents remain fundamentally episodic: they fail to retain, refine, and reuse experiences across tasks, repeatedly reconstructing context from scratch and reproducing similar mistakes. Even with memory support, they offer no remedy for the absence of a principled, task-agnostic extit{memory utility}, making them difficult to evaluate rigorously or generalize across agents and settings. To tackle these limitations, we introduce ours, a closed-loop framework for memory augmentation in SE agents. ours grounds memory utility in extit{validated downstream impact}, establishing utility as both a task-agnostic extbf{evaluation benchmark} and an annotation-free extbf{optimization signal}. Through complementary evaluation on extit{single-episode} and extit{cross-episode} memory augmentation, results demonstrate that ours consistently improves SE agents across settings, achieving absolute gains of up to parrow5.25% in success rate and parrow4.63% in resolve efficiency, while substantially reducing computational cost by geq9.79%. Our project page: href{https://xhguo7.github.io/MemOp/}{https://xhguo7.github.io/MemOp/}.

Source: arXiv cs.AI | 2026-06-06