The paper introduces a self-evolving search index that dynamically adapts document key representations to the varying demands of different retrieval environments. This approach addresses the challenge that fixed optimization strategies cannot consistently maintain retrieval quality when effective index representations differ across contexts. By evolving the index, the method aims to expose more relevant knowledge within each document for improved information retrieval.