EvoOntology: A Self-Evolving Ontology Layer for Data Agents

The paper introduces EvoOntology, a self-evolving ontology layer packaged as an MCP server comprising schema, content, and tool layers to help data agents actively query and interact with heterogeneous data sources. It features an autonomous builder agent and a self-evolution loop that continuously refines the ontology via attribution-guided typed edits accepted only after backbone-conditional paired evaluations. Experiments across three data-agent benchmarks with four LLM backbones show that EvoOntology consistently outperforms strong baselines and semantic-layer approaches.

EvoOntology consistently outperforms strong baselines and existing semantic-layer approaches across three data-agent benchmarks.

The framework encapsulates the ontology as an MCP server featuring a schema layer, a content layer, and a tool layer.

A self-evolution loop continuously refines the ontology through attribution-guided typed edits accepted after backbone-conditional evaluation.