Research map / Probabilistic and scientific modeling / Causal discovery
CaTs and DAGs: Integrating Directed Acyclic Graphs with Transformers for Causally Constrained Predictions
Matthew James Vowels, Mathieu Rochat, Sina Akbari
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 |
|---|---|---|---|
| use CFCN mod (direct linear input connection per layer) instead of CFCN (ablation) | CFCN · Twins | eATE out of sample: .0102 -> .0065 | Improved |
| use CFCN mod instead of CFCN (ablation) | CFCN · Jobs | policy risk out of sample: .26 -> .27 | No clear effect |
| replace CaT with CFCN (masked MLP, no attention) (ablation) | CaT · Twins | eATE out of sample: .0133 -> .0102 | No clear effect |
| use correct causal graph versus misspecified graph (ablation) | CaT / CFCN · toy motivating example | eATE | Improved |
| impose DAG constraint versus unconstrained transformer/MLP/RF under covariate shift | CaT, CFCN · toy motivating example | robustness to covariate shift | Improved |
| correct causal graph versus non-causal graph (ablation) | CaT · toy motivating example | MSE | Hurt |
| use CaT with full questionnaire items versus aggregated scores | CaT · COVID attachment-depression, fearful | effect size: .05 -> .01 | Mixed |
| shuffle variables in CaT (ablation) | CaT · predictive metrics | MSE | Hurt |
Closest papers by meaning
- Differentiable Structure Learning with Ancestral Constraints (ICML 2025 · similarity 0.88)
- Directed Graph Grammars for Sequence-based Learning (ICML 2025 · similarity 0.86)
- Effective Causal Discovery under Identifiable Heteroscedastic Noise Model (AAAI 2024 · similarity 0.86)
- Causal Concept Graph Models: Beyond Causal Opacity in Deep Learning (ICLR 2025 · similarity 0.86)
- A Recipe for Causal Graph Regression: Confounding Effects Revisited (ICML 2025 · similarity 0.86)
- ProDAG: Projected Variational Inference for Directed Acyclic Graphs (NeurIPS 2025 · 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.
