AI Paper Atlas

Research map / Probabilistic and scientific modeling / Treatment effects

Gradient-Based Causal Tree Ensembles: A Backbone Architecture for Heterogeneous Treatment Effects

Yusuke Kano, Jeremy Paul Voisey, Mihaela van der Schaar

ICML 2026 · Probabilistic and scientific modeling › Treatment effects · approach: Meta-learners and partial identification

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 changedWhereMeasuredEffect
replace fully-connected layers with GRACE tree layersTARNet · Toy dataset, k=100 featuresCATE RMSE: 0.135 -> 0.024Improved
use GRACE instead of Causal ForestCausal Forest · Toy dataset, k=5CATE RMSE: 0.041 -> 0.024Improved
add ensemble plus sparse feature selection to FCNN layers (FCNN_ES) (ablation)TARNet-like FCNN · Synthetic binary treatment (3 DGPs)RMSEImproved
add tree routing (T) to ensemble and sparse selection (ablation)FCNN_ES · Interaction-based DGPRMSEImproved
add tree routing (T) (ablation)FCNN_ES · FCNN-based DGPRMSEImproved
add tree routing (T) (ablation)FCNN_ES · Tree-based DGPRMSENo clear effect
use GRACE instead of baselines across sample sizes 500 to 10,000neural and tree baselines · Synthetic datasets (25 features)RMSEImproved
use GRACE instead of baselines on semi-synthetic dataneural and tree baselines · IHDP, ACIC 2016, TCGA, News, GlucoseRMSEImproved

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy