#software testing
Nov 2026
A knowledge distillation-based approach for line-level software defect prediction
DisTillMT is a collaborative learning approach that coordinates file-level and line-level SDP on a unified graph that significantly outperforms ten advanced line-level SDP methods by achieving accuracy gains of 20.35%–91.62%.
Fa Zhong, Rui-Lian Zhao, Ze-Ji Li et al.
· Empirical Software Engineeri... · 0 citations