["Journal of Behavioral Decision Making, Volume 39, Issue 4, October 2026. ", "\nABSTRACT\nThis study examined the dissociable effects of cognitive control and reward sensitivity on adolescents' decision‐making under ambiguity and risk by using the Iowa Gambling Task (IGT). Computational modeling was used to illuminate the underlying mechanisms. A two‐wave longitudinal design was employed with 339 early adolescents (M = 13.38 years at baseline, T1) who completed three cognitive tasks—the Stroop (cognitive control), Monetary Incentive Delay (reward sensitivity), and IGT (uncertain decision‐making) tasks—all at T1 and one‐year follow‐up (T2). IGT performance was decomposed into latent cognitive parameters (learning rate, outcome sensitivity, loss aversion, and response consistency) using the Prospect Valence Learning Delta model separately for the early stage (first 40 trials, predominantly ambiguous) and late stage (last 60 trials, predominantly risky) of the IGT. Controlling for sex and age, cross‐lagged panel models revealed that poorer cognitive control longitudinally predicted better late‐stage IGT performance (β = 0.13). At the cross‐sectional level, reward sensitivity was positively associated with early‐stage IGT performance (r = 0.15). Computational modeling revealed that reward sensitivity positively longitudinally predicted greater loss aversion during the early stage of the IGT (β = 0.13), whereas poorer cognitive control negatively longitudinally predicted outcome sensitivity during the late stage of the IGT (β = −0.13). These findings confirm a developmental dissociation in adolescent's decision‐making under uncertainty. Reward sensitivity facilitates effective encoding of trial feedback during the early stage of the IGT, whereas cognitive control underpins adaptive late‐stage IGT performance via stable value‐based choices. Computational modeling uncovered latent decision processes obscured by traditional behavioral metrics, advancing the understanding of dual‐system dynamics in adolescent decision‐making and informing targeted interventions such as cognitive control training for risk‐laden contexts and emotion regulation for ambiguous decision environments.\n"]