arXiv cs.AI· Paul Sevellec, Elisa Fromont, Romaric Gaudel, Laurence Roze, Paul Viallard, Laetitia Houeix·· 14 小时前AI 评分20
CounterFlow：面向视觉反事实解释的目标条件流匹配框架
CounterFlow: Target-Conditioned Flow Matching for Visual Counterfactuals
AI 导读
CounterFlow 是一个基于条件流匹配的视觉反事实解释框架，通过目标条件向量场将源图像直接引导至用户指定的目标类别。它在标准基准上实现了视觉真实感与目标类别切换的 SOTA 表现，且计算开销显著低于传统 GAN 或扩散模型架构。该方法直接在原始像素空间工作，更适合真实场景部署。
正文 · 原文
Abstract:In eXplainable AI, counterfactual explanations have gained significant traction by offering users direct, intuitive, and actionable insights into model decisions. We introduce CounterFlow, a novel framework based on Conditional Flow Matching that generates visual counterfactual explanations by directly steering a source image toward a user-specified target class. By leveraging target-conditioned vector fields, CounterFlow identifies highly realistic semantic modifications while requiring substantially lower computational overhead than traditional GAN- or diffusion-based architectures. Empirical evaluations on standard benchmarks demonstrate that CounterFlow achieves state-of-the-art performance in both visual realism and target class switching. By working directly in the raw pixel space, CounterFlow offers a more practical solution for real-world deployment.
| Subjects: | Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV) |
| Cite as: | arXiv:2610.35679 [cs.LG] |
| (or arXiv:2610.35679v1 [cs.LG] for this version) | |
| https://doi.org/10.48550/arXiv.2610.35679 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Paul Sevellec [view email]
[v1]
Mon, 26 Oct 2026 16:18:02 UTC (1,321 KB)
来源：arXiv cs.AI · arxiv.org