Get Started
Home
Topics
Search
Library
Research questionHow can we generate realistic, minimally edited visual counterfactuals without inheriting classifier shortcuts?Image counterfactuals can reflect a classifier’s shortcut features or calibration failures instead of genuine class differences. The central difficulty is changing class-salient visual factors while keeping the result realistic and minimally altered.
AI
Computer Vision
Health
Image & Video Processing
Image Generation
Machine Learning
Research Paper
Latest papersRecent research connected to this question, newest first.Counterfactual Contrastive AnalysisThe source studies classifier-free counterfactual generation from two class-specific image datasets, including medical imaging, using StyleGAN2 feature-space synthesis. Evidence comes from three medical imaging datasets and compares counterfactual generation quality with existing approaches.research paper · Sep 3, 2026
Related questions
How can medical-image counterfactuals reveal a classifier’s decisions without inheriting a generator’s biases?How can counterfactual image edits be audited for a shared latent explanation across regions?How should we choose graph counterfactual explainers when small edits, realism, coverage, and quality conflict?How can we reduce fine-tuning-induced shortcut reliance on underrepresented groups without retraining or group labels?