Accurately aggregating uncertain numerical beliefs is a critical challenge in collective intelligence and forecast aggregation, particularly in complex multi-agent settings where individual competencies may vary. This paper introduces ENEA: Epistemic Numerical Estimate Aggregation, a novel framework that extends the classic Condorcet Jury Theorem to the aggregation of continuous numerical beliefs. This framework generalizes Voting for Bins, a method designed to aggregate imprecise probabilistic beliefs. ENEA provides novel probabilistic guarantees for identifying the true underlying value within a binned range, by strategically transforming numerical estimates into votes and leveraging individual agent competencies. To address the inherent heterogeneity of agents in multi-agent systems and to enhance the practical applicability and performance of numerical aggregation, we enhance ENEA by developing an optimally weighted variant. For this weighted approach, we derive success probability bounds that are specifically applicable to heterogeneous agents under optimal weights. Through one-shot simulations, we demonstrate ENEA’s robust performance across different data distributions and its applicability to scenarios involving correlated beliefs.