Offline A/B Testing of Slate Recommendation Systems with LLMs: Reducing the Dependency on Pre-Collected User Interaction Data
A validation protocol is introduced measuring the alignment of synthetic preferences with classical RecSys metrics and their compliance with preference axioms, and used to characterise how LLM pre-training and configuration affect slate preference articulation.
Baptiste Bonin, M. Heuillet, Audrey Durand
· 0 citations