Longitudinal causal effect estimation from observational data is complicated by time-varying confounding and treatment-confounder feedback loops.
Pretrained transformer estimates longitudinal causal effects for dynamic treatment policies
PICLE-PFN predicts conditional average potential outcomes across datasets and horizons in one forward pass, without retraining or hyperparameter tuning.
Big Tech
Wenxin Chen · Kyra Gan · Mingyu Lu · Zilin Jing · Fei Wang · Hoifung Poon · +2 more
Cornell University · University of Washington · Columbia University · Recursion Pharmaceuticals, Inc. · Microsoft Research
Research Digest··2 min read
The authors introduce PICLE-PFN, the first foundation model for longitudinal causal effect estimation that supports both static and dynamic treatment policies.
The Daily Edition
The front page in your inbox each weekday morning: the lead, the picks, and why each one earned its place.