A generative intervention model maps perturbation features to distributions over atomic interventions in a jointly learned causal model, enabling out-of-distribution prediction with mechanistic insight.
Differentiable causal discovery from interventional data
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2024 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Generative Intervention Models for Causal Perturbation Modeling
A generative intervention model maps perturbation features to distributions over atomic interventions in a jointly learned causal model, enabling out-of-distribution prediction with mechanistic insight.