Coupled or Independent? Fault Coupling in Agentic Medical Vision Systems

概要

Vision-language agents equipped with tools expose intermediate outputs for inspection and comparison with the model's reasoning, which may offer safety benefits over single-shot models. This paper tests that potential benefit when the sensor reading itself is untrusted, so that the tool and reasoning channels share a common visual fault. We consider a dermatology agent based on segmentation, contrastive retrieval, and an iterative vision-language reasoner and define a coupling taxonomy that distinguishes single-channel errors from joint errors that agree or disagree. On PAD-UFES-20, we show that typographic injection produces detectable directional retrieval capture, whereas physical corruption does not. Crop-first retrieval preserves the payload at a readable scale and produces 55% joint wrong agreement across all cases (71% of faulty cases); full-image retrieval produces 35% (43%). Under a binary decision, if both channels are wrong, their predictions must agree, so hard-label channel comparison cannot detect joint failures, limiting disagreement recall to 0.29 (crop-first) versus 0.57 (full-image). The direction of the order effect is reproduced on the independent HIBA test set.

参考リンク

• Paper ↗

リンク先のライセンス・再現性・データ条件をご確認ください。自動分類や実装の目安は保証ではありません。

実装を始めるプロンプト

参考リンクを確認したうえで、次の内容をコピーして利用できます。

# 論文・記事
Coupled or Independent? Fault Coupling in Agentic Medical Vision Systems

# 概要
Vision-language agents equipped with tools expose intermediate outputs for inspection and comparison with the model's reasoning, which may offer safety benefits over single-shot models. This paper…

# 原典
https://arxiv.org/abs/2610.40913

上記の手法を実装してください。実行前に依存関係とライセンスを確認してください。