---
title: "ICON 分解：用于模型审计的深度表示多变量概念级解释"
source: "arXiv AI/CL/LG（论文）"
source_tier: "t2"
url: "https://arxiv.org/abs/2608.26083v1"
permalink: "https://feed.astralor.com/items/itm_2dfc4cd11607bc388d85924a"
published: "2026-08-26T17:47:49.000Z"
score: 75
channel: "ai"
tags: ["ICON Decomposition","医学影像","捷径学习","概念可解释性","模型审计"]
---

# ICON 分解：用于模型审计的深度表示多变量概念级解释

## AI 点评

> 这篇论文切中可解释性方法的常见误区，把概念从'是否可解码'升级到'独特贡献'，思路清晰。对做模型审计和医疗 AI 的人来说，ICON 可能成为标准工具箱里的一件新工具。

## 速览

ICON Decomposition 是一种概念级可解释性方法，用于审计深度模型的捷径学习。与传统方法孤立检测概念不同，ICON 在控制所有其他概念和结果变量后，量化每个概念对网络层方差的独特贡献，避免相关性被误判为模型依赖。在合成数据上，其准确性优于七种基线；在皮肤病变和脑成像模型上，ICON 成功识别模型真正依赖的概念，并通过重训练和分布外测试验证了稀疏解释。

## 关键点

- 提出 ICON 分解方法，在多变量条件下量化概念对层方差的解释力，避免孤立评估的偏差。
- 在合成数据上，ICON 恢复概念重要性的准确率优于七种基线方法。
- 在皮肤病变和脑成像模型上，ICON 识别出模型真正依赖的稀疏概念，并通过重训练和分布外测试验证。
- 可量化未被给定概念解释的表示部分，为模型审计提供更全面的视角。

## 全文

Deep neural networks often exploit spurious associations in their training data, a failure known as shortcut learning. Concept-based explainability methods screen for shortcuts by testing whether concepts such as a patient's sex or scanner settings can be decoded from a network layer. Because each concept is evaluated in isolation, these methods can mistake correlations between concepts as evidence that the model uses them. We introduce ICON decomposition, which instead quantifies how much of a layer's variance each concept explains after accounting for all other concepts and the outcome. On synthetic data with known ground truth, ICON recovers concept importance more accurately than seven alternative baseline methods. On skin-lesion and brain-imaging models, it isolates the concepts on which a model genuinely relies, quantifies the representation unexplained by any of the supplied concepts, and yields sparse explanations that we validate by retraining and out-of-distribution testing.

---

引用前请核对原文;本文由星际流动整理:https://feed.astralor.com/items/itm_2dfc4cd11607bc388d85924a