How much harder is bandit learning when the action set is unbounded? Surprisingly, only by a logarithmic factor — provided the loss is well-conditioned.
Spotlight NeurIPS 2025 Bandits and online learning
On the Statistical Cost of Open-Vocabulary Decision-Making
Cite this paper
On the Statistical Cost of Open-Vocabulary Decision-Making
@inproceedings{aydn2025open,
title = {On the Statistical Cost of Open-Vocabulary Decision-Making},
author = {Leila Aydın and Jihoon Park and Sasha Volkov and Renat Ostrovsky},
booktitle = {NeurIPS 2025},
year = {2025}
}