Large language models for repository-level repair usually reset context for every issue, even though real maintenance often unfolds as a stream of tickets within the same codebase. We study this sequential regime on a stratified 8 × 25 subset of SWE-Bench Pro (eight projects, twenty-five instances each), running instances in commit-time order within each project so later repairs can draw on earlier trajectories. We present TwinMem-Agent, which pairs a test-patch-validate loop with episodic memory (grounded demonstrations) and semantic memory (LLM-distilled notes). After each issue, a memory module compacts the trajectory, merges updates under a fixed budget, and retrieves exemplars and guidelines at inference time, all without fine-tuning model weights. Using DeepSeek-V3.2 as the shared backbone and identical Dockerized instances, we compare TwinMem-Agent to mini-swe-agent on Pass@1 and average inference cost, and ablate the memory module on 25 Ansible instances measuring Pass@1 and RSR. TwinMem-Agent achieves 89.1% Pass@1 versus 60.0% for the baseline at roughly 18× lower cost ($0.0044 vs. $0.0798). Removing the memory module reduces Pass@1 by 8 pp and RSR by 12 pp, with the larger RSR drop confirming that accumulated cross-issue experience most directly benefits bug reproduction and cascades into higher patch success.
Benzhao Wu, Zhiwen Luo, Fangwen Mu et al.· SIGSOFT FSE Companion· 0 citations
The rapid advancement of LLMs has opened new opportunities in automated software engineering, driving progress in code understanding, agent-based workflows, and productivity tools. However, existing code intelligence systems have largely sidelined the end-users they aim to serve—the developers themselves. Developers exhibit substantial heterogeneity across multiple dimensions: coding style, toolchain preferences, domain-specific expertise, and problem-solving strategies. Failing to account for these individual differences directly compromises both the effectiveness of code intelligence and the likelihood of its adoption. For example, a senior architect and a junior engineer ask: "Describe the authorization module." Without personalized context, the system produces a uniform response—verbose for the expert, incomprehensible for the novice. This gap motivates a fundamental shift: from one-size-fits-all to one-size-fits-one code intelligence. A developer's dynamic in-IDE behaviors—code authoring patterns, navigation pathways, debugging trajectories—implicitly encode a rich representation of their competencies and habits. If captured and interpreted systematically, these signals can enable Personalized Code Intelligence, formalized as: [EQUATION] where P is the developer persona derived from IDE behaviors, injected alongside code context C and instruction ℐ.
Yuhong Liu, Yu Su, Zhipeng Peng et al.· SIGSOFT FSE Companion· 1 citation