Tag
This paper investigates the interpretability of DAPF-based models for dementia detection, revealing that while DAPF achieves strong performance, its token-level explanations lack faithfulness.