Behind the [MASK]: Disentangling Representation and Faithfulness in DAPF-Based Dementia Detection

2026-08-25

AI Summary

研究通过分析基于DAPF框架的痴呆症检测模型，解决了低资源条件下非侵入性痴呆筛查的问题，但发现该模型虽表现优异却缺乏忠实的词级解释。

Abstract

Spoken-language analysis via prompt-based domain-adaptive models is a promising direction for low-resource, non-invasive dementia screening, but such models remain internally opaque. We study the interpretability of the Domain-Adapted models via Prompt-based Fine-tuning (DAPF) framework, which casts dementia detection as diagnosis-related masked-token prediction. We interpret DAPF and strong baselines using a variety of probing and analysis techniques, finding that DAPF achieved the best overall performance (accuracy=0.83 and macro-F1=0.83) with diagnosis most recoverable from its [MASK] representation. However, this representational advantage did not extend to token-level explanation faithfulness. DAPF attributions primarily reflected language task vocabulary, discourse markers, and transcription artifacts, with perturbation tests showing weak or negative effects. This suggests that its masked-token interface determines diagnosis information without producing faithful token-level explanations.

Problem

• Dementia Detection
• Interpretability
• Faithfulness
• Masked-token Prediction
• Domain-Adapted Models

Innovation

• Domain-Adapted models
• Prompt-based Fine-tuning
• Interpretability
• Masked-token prediction
• Faithfulness