CAFD: Concept-Aware DNN Fault Detection using VLMs

In three linesCAFD is a DNN fault detection method combining model signals, distance features, and a novel Concept Failure Ratio (CFR) leveraging Vision-Language Models. Evaluated on ImageNet and three models, CAFD outperforms 5 baselines with average 18.3% improvement in Fault Detection Rate.

Summary generated by Claude — human-verified