Search papers, labs, and topics across Lattice.
The paper introduces SkillBrew, a framework for curating skill banks for retrieval-augmented LLM agents by formulating skill bank curation as a constrained multi-objective optimization problem, balancing utility, diversity, and coverage. SkillBrew uses a bi-level propose-then-verify loop to achieve Pareto-aware optimization under a utility constraint. Experiments on public benchmarks demonstrate the benefits of principled skill bank curation over append-only approaches for building more effective LLM agents.
LLM agents can be significantly improved by *removing* redundant and outdated skills from their skill banks, not just adding more.
Retrieval-augmented LLM agents increasingly rely on curated skill banks: collections of reusable textual principles that guide decision making on complex tasks. Existing approaches typically expand these banks in an append-only fashion, continuously adding new skills without removing redundant, outdated, or harmful ones, resulting in inefficient and poorly curated repositories. In this paper, we formulate the skill bank curation as a constrained multi-objective problem: a desirable bank must be useful for the agent, diverse in its content, and provide good coverage of the query distribution. To this end, we introduce SkillBrew, a multi-objective curation framework that formalizes skill bank curation as Pareto-aware optimization under a utility constraint, and solves it via a bi-level propose-then-verify loop. We evaluate our approach on two public benchmarks. Our findings suggest that treating skill banks as objects of principled curation, rather than ever-growing append-only logs, is an important step toward building self-improving LLM agents.