Task: 4-arm restless bandit
Human participants choose among four actions each trial and receive a scalar reward. Actions are encoded as integers 1–4.
This is a restless bandit-style sequential decision task used for evaluating agents on same-subject continuation and related behavioral prediction settings defined by the challenge.