Elliot Creager

Papers

Total Citations

H-Index

About

Elliot Creager is a researcher whose work bridges reinforcement learning, robotics, and fluid simulation. His key research areas include causal inference in dynamic systems, data augmentation for control, and physics-based machine learning. Creager’s major contribution is the development of **counterfactual data augmentation using locally factored dynamics** (2020, 11 citations), a method that exploits sparse interactions between subprocesses in robotic control and RL to generate realistic, causally grounded training data. This work addresses a critical challenge in reinforcement learning: learning robust policies from limited interactions with complex environments. More recently, Creager has ventured into **learning-based fluid simulation** with **SurfsUp** (2023, 2 citations), which models fluid mechanics on novel surfaces—a task vital for design, graphics, and robotics. By enabling fast, differentiable simulators that generalize to unseen geometries, this work pushes the boundaries of physics simulation. Though his citation counts are modest, Creager’s contributions are notable for their conceptual rigor and practical relevance, particularly in making RL and simulation more sample-efficient and generalizable. His research exemplifies a thoughtful integration of causal reasoning and machine learning for real-world control problems.