UniER: A Unified Benchmark for Item-level and Path-level Exercise Recommendation

In three linesUniER is a unified benchmark for personalized exercise recommendation, comparing two paradigms: ILER (item-level) and PLER (path-level). The framework introduces Weighted Cognitive Gain (WCG) metric and evaluates 18 methods across 9 datasets. Results show systematic dominance of PLER and reveal ILER's pedagogical failures under extreme sparsity and noise.

Summary generated by Claude — human-verified