arXiv cs.AI· Haoran Jin, Defu Lian, Xiting Wang, Hong Xie·· 14 小时前AI 评分34
MUSE：面向科学编程智能体的多级记忆与选择性运行时注入框架
MUSE: Multi-Level Memory with Selective Runtime Injection for Scientific Coding Agents
AI 导读
MUSE 是一个面向科学编程智能体的记忆管理框架，从历史轨迹构建全局、方法级和实现级三层记忆，初始化时注入全局与问题匹配的方法级记忆，执行中按当前交互检索实现级记忆，并由 LLM 门控决定是否注入审查建议。在数值问题求解、kernel 优化和科学软件修复四类基准上，MUSE 在所有评测指标上均优于检索与记忆基线，消融实验支持多级设计与门控注入的有效性。代码已开源。
正文 · 原文
Abstract:Historical experience can guide scientific coding at multiple levels, but task-level retrieval may miss implementation lessons that become relevant during execution. We introduce MUSE, a memory management framework combining multi-level memory with selective runtime injection for scientific coding agents. MUSE builds global, approach-level, and implementation-level memory from historical trajectories. At initialization, the agent receives global memory and problem-matched approach-level memory. During execution, current interactions are used to retrieve implementation-level memory. An LLM gate selects relevant experience and generates specific review advice, or skips injection. Across four benchmarks spanning numerical problem solving, kernel optimization, and scientific software repair, MUSE outperforms retrieval and memory baselines on all evaluation metrics. Ablations support the multi-level design and gated injection, while transfer experiments show that memory remains useful across kernel benchmarks. These results show that multi-level memory and selective runtime injection enable effective reuse of experience in scientific coding. Code is available at this https URL.
| Comments: | 17 pages, 8 figures. Under review |
| Subjects: | Machine Learning (cs.LG); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2610.38332 [cs.LG] |
| (or arXiv:2610.38332v1 [cs.LG] for this version) | |
| https://doi.org/10.48550/arXiv.2610.38332 arXiv-issued DOI via DataCite |
Submission history
From: Haoran Jin [view email]
[v1]
Wed, 28 Oct 2026 06:32:51 UTC (259 KB)
来源：arXiv cs.AI · arxiv.org