Reward associations can shape attentional priorities, but the conditions under which such learning persists and generalizes remain unclear. We examined how learned item-reward associations influence behavior within (Experiment 1) and between (Experiment 2) conflict tasks when rewards are removed. In both experiments, participants performed a color-word Stroop task, responding to the color while ignoring the word. Certain colors were associated with high (90%) or low (10%) reward probability. In Experiment 1, we compared learning with explicit pre-cues signaling reward availability to learning through feedback alone. Both groups responded faster to high-reward colors, and the pre-cued group additionally showed enhanced focus (reduced Stroop effect) on reward trials. However, a critical dissociation emerged once rewards were removed: only participants who learned without pre-cues maintained their performance advantage for previously high-reward colors. Thus, explicit pre-cues promote immediate, context-dependent control, whereas learning without pre-cues fosters durable, automatic attentional biases that persist beyond training. Experiment 2 tested whether item-reward associations transfer across tasks. After Stroop training without pre-cues, participants completed a color flanker task (responding to the center circle’s color while ignoring outer circles), which shared perceptual features (colors) with the Stroop task but involved no rewards. Despite changes in task structure, stimuli, and conflict type, participants continued responding faster to previously high-reward colors. These findings reveal that item-reward associations can produce lasting, transferable effects on attention, highlighting a mechanism through which reward shapes attention and behavior in conflict tasks beyond the immediate training context.