Calibration-Gated LLM Pseudo-Observations for Online Contextual Bandits

This research introduces a novel calibration-gated mechanism using LLM pseudo-observations to solve the cold-start problem in contextual bandits, significant...

Level: advanced

By Maksim Pershin, Ivan Golovanov, Pavel Baltabaev, Natalia Trankova

Category: research