InferenceEvolve : Automated Causal Effect Estimators through Self-Evolving AI

Can Wang, Hongyu Zhao, * · 2026

Read InferenceEvolve : Automated Causal Effect Estimators through Self-Evolving AI by Can Wang, Hongyu Zhao, * (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2604-04274.txt.

Description

Wang, Zhao and Chen build InferenceEvolve, a system that uses large language models to invent and refine causal inference methods by evolution. Choosing the right causal method is hard, given tangled statistics and messy real data. Across standard benchmarks their evolved estimators beat established baselines. In a recent community contest their best entry sat on the Pareto frontier across two metrics, against 58 human entries. The team also built proxy goals for settings without semi-synthetic outcomes, and found agents gradually uncovering strategies fitted to hidden data mechanisms.

Catalog record

CREATORCan Wang, Hongyu Zhao, *
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE97163
LICENSEhttp://creativecommons.org/licenses/by-sa/4.0/
RIGHTSCC BY SA
SOURCEarXiv

Source: https://arxiv.org/abs/2604.04274

Browse the full library · About the Tower of Babel