AI Paper Atlas

Research map / Probabilistic and scientific modeling / Treatment effects

Causal Representation Learning with Optimal Compression and Complex Treatments

Wanting Liang, Haoang Chi, Zhiheng Zhang

ICML 2026 · Probabilistic and scientific modeling › Treatment effects · approach: Benchmarking treatment effect models

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
add any multi-treatment balancing strategy (Pairwise, OVA, Aggregation)Base Model · Semi-synthetic, K=4PEHE: 0.796 -> 0.711 (OVA), 0.727 (Pairwise), 0.722 (Agg-T)Improved
use One-vs-All instead of Pairwise balancing at small K (ablation)balancing strategy · Semi-synthetic, K=4PEHE: 0.727 -> 0.711Improved
use Treatment Aggregation (HSIC) instead of Pairwise balancing at small K (ablation)balancing strategy · Semi-synthetic, K=4PEHE: 0.727 -> 0.722No clear effect
use strong regularization alpha=5.0 with Pairwise balancing at large K (ablation)Pairwise balancing · Semi-synthetic, K=20PEHE: PEHE above 1.3Hurt
use Treatment Aggregation instead of Pairwise as K grows to 20 (ablation)balancing strategy · Semi-synthetic, K=20PEHE and stability: Agg-T PEHE about 1.0 across all alpha vs Pairwise above 1.3 at alpha=5.0Improved
tune alpha per strategy (ablation)balancing strategies · Semi-synthetic, K=4best alpha: OVA 0.1, Aggregation 0.5, Pairwise 5.0Improved
use generative Multi-Treatment CausalEGM instead of the unadjusted baseCausalEGM · UCI DigitsPEHE: 0.79 -> 0.65Improved
use generative CausalEGM instead of discriminative OVAbalancing strategy · UCI DigitsPEHE: 0.24 vs 0.65Hurt

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