ArXiv: Knowledge Compounding — Empirical Economics of Self-Evolving Agent Knowledge Wikis Under Agentic ROI Framework
arXiv·medium signal
Paper arXiv:2604.11243 introduces knowledge compounding as a measurable economic concept for LLM agents, analyzing how self-evolving knowledge wikis accumulate value over iterative agent runs. The paper proposes an Agentic ROI Framework for quantifying the returns on agent-maintained knowledge infrastructure. Directly relevant to builders running persistent agent systems with self-improving knowledge bases.