The long-standing Yerkes-Dodson law proposes an inverted-U-shaped relationship between arousal and cognitive performance, yet empirical support in humans has been mixed. Recent studies using pupil-linked arousal have shown that spontaneous fluctuations in arousal relate to perceptual performance, and that pharmacologically elevating arousal shifts the peak of this curve without disrupting its overall inverted-U shape. It remains unclear, however, whether the same mechanism holds when arousal is increased naturally, such as through physical exercise. This study tested how moderate-intensity exercise influences arousal and perceptual decision-making, and whether exercise alters the arousal-performance relationship, and explored how arousal and performance change over time within a task block. Twenty-six participants completed an auditory detection task after either moderate cycling or rest, while pupil size indexed arousal. Exercise reliably elevated physiological arousal, but behavioral performance (reaction time and sensitivity) did not differ between conditions. Mixed-effects models provided no evidence for an inverted-U-shaped relationship: instead, higher pupil-linked arousal was linearly associated with reduced sensitivity and slower responses in both exercise and rest. Temporal analyses further showed the expected decline in arousal over each block, with only a modest coupling between moment-to-moment arousal and performance. Together, these findings suggest that moderate exercise elevates arousal but does not shift arousal-performance dynamics, highlighting important differences between naturalistic and pharmacological arousal modulation.