A frozen in-context learning model, conditioned on bootstrap resamples of the interaction history and an arm-context feature map, selects actions and beats several contextual bandit baselines in cumulative regret.
These methods enjoy strong theoretical foundations but are limited in expressiveness and can perform poorly under model misspecification (Lattimore & Szepesv´ari, 2017)
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2026 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Bootstrap-Conditioned Action Selection with Tabular Foundation Models
A frozen in-context learning model, conditioned on bootstrap resamples of the interaction history and an arm-context feature map, selects actions and beats several contextual bandit baselines in cumulative regret.