Experience-RAG: Agent-Oriented Pluggable Skill for Adaptive Retrieval Strategy Selection

arXiv·medium signal

Experience-RAG introduces a pluggable agent skill that learns which retrieval strategy works best for which task type — factoid QA, multi-hop reasoning, and scientific verification each get different retrieval pipelines automatically. Instead of a one-size-fits-all RAG pipeline, the system uses accumulated experience to orchestrate strategy selection per query. Directly applicable for builders running multi-purpose RAG agents where a single retrieval approach leaves performance on the table.