This paper defines the three challenges of learning-based set size estimation (generality, accuracy, and updatability) as the "Triangular Dilemma of Set Size Estimation" and proposes DistJoin, an efficient, distribution-based join set size estimator using a multi-autoregressive model. DistJoin separately utilizes the probability distributions of individual tables to estimate the join set size and achieves efficiency through Adaptive Neural Predicate Modulation (ANPM), a high-throughput distribution estimation model. Furthermore, we formally address the variance accumulation problem of existing similar approaches through variance analysis and effectively reduce variance through a selectivity-based approach. DistJoin is the first data-driven method to support both equi- and non-equi-joins, offering high accuracy and fast, flexible updates. Experimental results show that DistJoin achieves the highest accuracy, robustness to data updates, and generality compared to existing methods, while demonstrating comparable update and inference speeds.
