FiRe: Fixed-Noise Refinement for Visual Counterfactual Explanations

2026-08-09

AI Summary

This work addresses the limitations of existing DDPM-based visual counterfactual methods, which struggle to balance semantic editability and spatial control over long denoising trajectories and often rely on computationally expensive or low-quality estimates to obtain clean images for classifier guidance. To overcome these challenges, the authors propose FiRe, a framework that performs iterative optimization at a fixed noise level and leverages Pixel Mean Flow to directly predict clean images from noisy states. FiRe further incorporates dynamic dual masking, adaptive classifier guidance, and an early-stopping mechanism to enable efficient and precise local editing. Evaluated across five tasks on three datasets, FiRe achieves approximately 3× faster inference and 8× fewer FLOPs than the strongest baseline while maintaining or improving the quality of counterfactual explanations.

Abstract

Visual counterfactual explanations aim to change classifier decisions through realistic and localized edits while preserving decision-irrelevant content. Existing DDPM-based methods typically perform classifier-guided editing along a long reverse denoising trajectory. The changing noise levels make semantic editability and spatial control difficult to balance, and the editable state is noisy, whereas the target classifier is trained on clean images. As a result, these methods require either costly recursive denoising or low-quality one-step estimates to obtain classifier-facing clean images. We propose FiRe, a Fixed-noise Refinement framework for visual counterfactual explanations. Rather than following a reverse denoising trajectory, FiRe maps the input to a fixed noise level and iteratively refines the noisy state at that level. To provide clean images for classifier guidance, FiRe first adapts Pixel Mean Flow to visual counterfactual explanation, enabling direct clean-image prediction from noisy states. To make fixed-noise refinement produce minimal and localized counterfactual edits, FiRe introduces three FiRe-specific controls: a dynamic dual-mask strategy, adaptive guidance, and early stopping, which determine where edits accumulate, which changes become visible, and when refinement stops. Experiments on five tasks across three datasets show that, compared with the strongest recent baseline, FiRe achieves about 3$\times$ faster online inference and 8$\times$ fewer FLOPs while obtaining comparable or state-of-the-art counterfactual quality.

Problem

• visual counterfactual explanations
• DDPM
• classifier guidance
• denoising trajectory
• clean image prediction

Innovation

• Fixed-noise Refinement
• Visual Counterfactual Explanation
• Pixel Mean Flow
• Dynamic Dual-Mask
• Adaptive Guidance