Research map / Image generation / Text-to-image diffusion
Causal-Adapter: Taming Text-to-Image Diffusion for Faithful Counterfactual Generation
Lei Tong, Zhihua Liu, Chaochao Lu, Dino Oglic, Tom Diethe, Philip Alexander Teare, Sotirios A. Tsaftaris, Chen Jin
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add prompt-aligned injection (PAI) (ablation) | Causal-Adapter · CelebA validation | average F1: +11.6% F1 | Improved |
| add conditioned token contrast (CTC) (ablation) | Causal-Adapter · CelebA validation | CLD: 0.317 -> 0.310 | Improved |
| add conditioned token contrast (CTC) (ablation) | Causal-Adapter · CelebA validation | FID: 8.453 -> 8.643 | Hurt |
| add attention guidance for localized editing | Causal-Adapter · CelebA | FID: 8.152 -> 5.213 | Improved |
| add causal mechanisms (SCM) to Causal-Adapter (ablation) | Causal-Adapter · Pendulum do(p) | pendulum MAE: 0.159 -> 0.014 | Improved |
| add causal mechanisms (SCM) to Causal-Adapter (ablation) | Causal-Adapter · Pendulum do(l) | light MAE: 0.173 -> 0.041 | Improved |
| use correct causal graph SCM2 instead of incorrect SCM1 (ablation) | Causal-Adapter · Pendulum do(l) | light MAE: 0.062 -> 0.041 | Improved |
| inject ground-truth labels instead of causal mechanisms | Causal-Adapter · Pendulum do(p) | pendulum MAE: 0.014 -> 0.013 | No clear effect |
| swap U-Net backbone for DiT backbone | Causal-Adapter · CelebA-HQ | eyeglasses F1: 99.26 -> 97.39 | No clear effect |
| add attention guidance for localized editing | Causal-Adapter · ADNI | FID: 9.130 -> 9.066 | No clear effect |
Closest papers by meaning
- Counterfactual Image Editing (ICML 2024 · similarity 0.91)
- Diffusion Counterfactual Generation with Semantic Abduction (ICML 2025 · similarity 0.91)
- Localizing and Editing Knowledge In Text-to-Image Generative Models (ICLR 2024 · similarity 0.90)
- Benchmarking Counterfactual Image Generation (NeurIPS 2024 · similarity 0.88)
- Visual Representation Learning through Causal Intervention for Controllable Image Editing (CVPR 2025 · similarity 0.88)
- Towards Safe Concept Transfer of Multi-Modal Diffusion via Causal Representation Editing (NeurIPS 2024 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
