HOBA Framework Uses Hierarchical RL and LLMs for Adaptive Ad Bidding

HOBA employs a three-tier hierarchical reinforcement learning framework to automate online advertising bidding. An LLM handles high-level hyperparameter reasoning via a reflection loop, a SARSA agent manages mid-level model selection, and a low-level expert pool executes specific bids.