Research map / Probabilistic and scientific modeling / Causal discovery
High Fidelity Image Counterfactuals with Probabilistic Causal Models
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Replace the baseline image mechanism with an exogenous-prior HVAE | Morpho-MNIST SCM · Morpho-MNIST test, do(y) | digit accuracy (%): 83.18 -> 99.49 | Helped |
| Use the latent-mediator HVAE for direct effects instead of the baseline | Morpho-MNIST SCM · Morpho-MNIST test, mix | intensity MAE: 9.61 -> 1.89 | Helped |
| Use the latent-mediator HVAE for total effects rather than direct effects | Morpho-MNIST SCM · Morpho-MNIST test, do(y) | digit accuracy (%): 99.47 -> 99.68 | Helped |
| Remove parent conditioning dropout (ablation) | MED_DE · Morpho-MNIST test, do(y) | digit accuracy (%): 99.47 -> 81.07 | Hurt |
| Remove parent conditioning dropout (ablation) | MED_TE · Morpho-MNIST test, do(y) | digit accuracy (%): 99.68 -> 94.28 | Hurt |
| Use parent conditioning dropout (ablation) | MED_DE · Morpho-MNIST test | bpd: .676 -> .682 | Hurt |
| Increase β from 1 to 3 (ablation) | EXO_DE · Morpho-MNIST test, do(y) | digit accuracy (%): 99.49 -> 99.73 | Helped |
| Increase β from 1 to 3 (ablation) | MED_TE · Morpho-MNIST test, do(y) | digit accuracy (%): 99.68 -> 99.79 | Helped |
| Fine-tune the image mechanism with counterfactual training | UK Biobank exogenous-prior HVAE · UK Biobank test, do(a) | age MAE (years): ↑ 0.004 | Hurt |
| Fine-tune the image mechanism with counterfactual training | MIMIC-CXR exogenous-prior HVAE · MIMIC-CXR test, do(a) | age MAE (years): ↓ 3.426 | Helped |
Closest papers by meaning
- Counterfactual Image Editing (ICML 2024 · similarity 0.92)
- Counterfactual Generative Modeling with Variational Causal Inference (ICLR 2025 · similarity 0.91)
- Towards Characterizing Domain Counterfactuals for Invertible Latent Causal Models (ICLR 2024 · similarity 0.91)
- Benchmarking Counterfactual Image Generation (NeurIPS 2024 · similarity 0.89)
- Diffusion Counterfactual Generation with Semantic Abduction (ICML 2025 · similarity 0.89)
- Exogenous Matching: Learning Good Proposals for Tractable Counterfactual Estimation (NeurIPS 2024 · similarity 0.87)
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.