EvoOntology: A Self-Evolving Ontology Layer for Data Agents
EvoOntology gives data agents a runtime ontology server that can build and revise its own semantic layer.
The paper frames the problem as an “agent-data gap” between natural-language agents and heterogeneous tables, files, and databases. Its ontology is exposed through an MCP server with schema, content, and tool layers. A builder agent constructs the ontology, then a self-evolution loop refines it using attribution-guided typed edits and paired evaluation. The authors report gains over strong baselines and other semantic-layer methods across three data-agent benchmarks and four LLM backbones. ArXiv · AI/CL/LG's note
The paper frames the problem as an “agent-data gap” between natural-language agents and heterogeneous tables, files, and databases. Its ontology is exposed through an MCP server with schema, content, and tool layers. A builder agent constructs the ontology, then a self-evolution loop refines it using attribution-guided typed edits and paired evaluation. The authors report gains over strong baselines and other semantic-layer methods across three data-agent benchmarks and four LLM backbones. ArXiv · AI/CL/LG's note
score 5