Saved in:
Bibliographic Details
Main Authors: Zhu, JiaWei, Liu, ZiHeng
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2512.19349
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866909973533949952
author Zhu, JiaWei
Liu, ZiHeng
author_facet Zhu, JiaWei
Liu, ZiHeng
contents Hidden confounding remains a fundamental challenge in causal inference from observational data. Recent advances leverage Large Language Models (LLMs) to generate plausible hidden confounders based on domain knowledge, yet a critical gap exists: LLM-generated confounders often exhibit semantic plausibility without statistical utility. We propose VIGOR+ (Variational Information Gain for iterative cOnfounder Refinement), a novel framework that closes the loop between LLM-based confounder generation and CEVAE-based statistical validation. Unlike prior approaches that treat generation and validation as separate stages, VIGOR+ establishes an iterative feedback mechanism: validation signals from CEVAE (including information gain, latent consistency metrics, and diagnostic messages) are transformed into natural language feedback that guides subsequent LLM generation rounds. This iterative refinement continues until convergence criteria are met. We formalize the feedback mechanism, prove convergence properties under mild assumptions, and provide a complete algorithmic framework.
format Preprint
id arxiv_https___arxiv_org_abs_2512_19349
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle VIGOR+: Iterative Confounder Generation and Validation via LLM-CEVAE Feedback Loop
Zhu, JiaWei
Liu, ZiHeng
Artificial Intelligence
Machine Learning
I.2.11; I.2.7; G.3
Hidden confounding remains a fundamental challenge in causal inference from observational data. Recent advances leverage Large Language Models (LLMs) to generate plausible hidden confounders based on domain knowledge, yet a critical gap exists: LLM-generated confounders often exhibit semantic plausibility without statistical utility. We propose VIGOR+ (Variational Information Gain for iterative cOnfounder Refinement), a novel framework that closes the loop between LLM-based confounder generation and CEVAE-based statistical validation. Unlike prior approaches that treat generation and validation as separate stages, VIGOR+ establishes an iterative feedback mechanism: validation signals from CEVAE (including information gain, latent consistency metrics, and diagnostic messages) are transformed into natural language feedback that guides subsequent LLM generation rounds. This iterative refinement continues until convergence criteria are met. We formalize the feedback mechanism, prove convergence properties under mild assumptions, and provide a complete algorithmic framework.
title VIGOR+: Iterative Confounder Generation and Validation via LLM-CEVAE Feedback Loop
topic Artificial Intelligence
Machine Learning
I.2.11; I.2.7; G.3
url https://arxiv.org/abs/2512.19349