Databricks research shows multi-step agents consistently outperform single-turn RAG when answers span databases and documents

To confirm this is an architecture problem rather than a model quality problem, Databricks reran published STaRK baselines ...
IEEE

Structured KV Cache Selection with RoPE Correction for Efficient Multi-Query Large Table Question Answering

Abstract: The application of Large Language Models to complex reasoning tasks over large-scale, structured data, such as enterprise tables, is becoming increasingly prevalent. However, this ...