Beliefs about relative academic performance may shape college major choice and explain gender gaps in STEM, but little causal evidence exists. To test whether these beliefs are malleable and salient enough to change behavior, I run a randomized experiment with 5,700 undergraduates across seven introductory STEM courses. Providing relative performance information shrinks gender gaps in biased beliefs substantially. However, students' course-taking and major choice are largely unchanged. If anything, initially overconfident men and women were discouraged by the intervention. Increasing female STEM participation may require more intensive or targeted intervention. [This paper was published in "Economics of Education Review" v97 Article 102479 2023.]