Language model performance improves with reusable plugins in harnesses
PluginRSI: Recursive Improvement of Agent Harnesses with Reusable Plugins
Artificial Intelligence
Summary
Making smart programs that guide language models to solve problems works better when the program is built from small, reusable pieces called plugins. The authors created a method called PluginRSI, which improves each plugin on its own and then mixes them to make better guiding programs. This approach helps the language model do better on different tasks like coding, using command lines, and answering questions, even when used with new models. Over time, PluginRSI builds a library of useful plugins that speeds up creating even better programs for new challenges.
What this means in practice
- •For software developers: Improve automation tools by combining reusable plugins that improve language model guidance for coding tasks.
- •For technical support teams: Build more effective command-line assistants by reusing optimized plugins to guide language model interactions.
Authors
Yaorui Shi, Yuchun Miao, Yuxin Chen, Jiayuan Zhang, Yueqing Sun, Xierui Song, Xiang Wang, An Zhang
Abstract
The harness surrounding a language model is a central determinant of agent performance. Recent methods optimize harnesses by searching over complete programs, where individual mechanisms are difficult to isolate and reuse. We introduce PluginRSI, which represents a harness as a composition of atomized plugins and organizes harness evolution around these plugins. Individual plugins are improved independently and accumulated in a shared library, then recombined into new harnesses at each iteration. PluginRSI improves over existing harness optimization methods across software engineering, command-line interaction, and question-answering tasks. The resulting harnesses retain their advantage when transferred to other solver models without further optimization. The evolved plugin library accelerates subsequent optimization from the initial harness, which helps faster and higher convergence on unseen tasks. These results show that accumulating reusable mechanisms provides an effective basis for continued harness improvement.