Renmin University open-sources FlashRSI, a unified self-improving agent framework
Instead of manually tuning a coding agent's prompts, skills and tool configs, teams can let FlashRSI run the propose-edit-evaluate-select loop and benchmark which changes to keep. Builders maintaining long-running agents — repeated file reads, forgotten post-fix tests, token waste — get an auditable changelog of every config revision. The open-source release lets practitioners pick a specific algorithm like GEPA or ACE per optimisation layer rather than rewriting one-off tuning scripts.
Renmin University (RUC) released FlashRSI as an open-source unified framework for recursive self-improvement (RSI) of AI agents. It automates the full optimisation loop: proposing improvements, editing agent configurations, running tasks, evaluating results, selecting better versions, and iterating. FlashRSI unifies 13 agent self-improvement algorithms, integrating OPRO, GEPA, ACE, SkillOpt, Meta-Harness and RRSI across prompt, skill, experience-memory and harness layers. Every round logs which prompts, skills, tools or config files were changed, along with candidate versions, execution traces, evaluation scores and resource consumption.
人大最新开源：FlashRSI，一个面向RSI的统一框架，让Agent自动优化自己的Prompt、Skill、工具、Harness
举个例子，比如一个Coding Agent经常重复读取文件、浪费 Token，或者修复Bug后忘记运行测试，你就需要反复修改Prompt、Skill、工具配置，再测试效果
FlashRSI把这套优化过程给自动化了 提出改进、修改Agent配置、运行任务、评估效果、筛选更优版本、继续迭代流程自动化
它统一了13种Agent自我改进算法，集成了OPRO、GEPA、ACE、SkillOpt、Meta-Harness、RRSI等，覆盖 Prompt、Skill、经验记忆以及Harness等不同优化层次
你可以通过真实任务和Benchmark判断哪些修改值得保留
每轮改进也都会记录Agent修改了哪些Prompt、Skill、工具或配置文件，保留候选版本、执行轨迹、评测分数和资源消耗
#FlashRSI #RSI #AIRSI
