# LLM-Augmented Causal Discovery: Probabilistic Fusion of Edge Existence and Orientation

arXiv: https://arxiv.org/abs/2608.27472 · HTML: https://mlsanity.com/paper/2608.27472/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 7 of its 17 references that are in the map.
- First posted on arXiv 19 Aug 2026.

## Builds on (most cited first, 7 of 7)

- [A survey of Bayesian Network structure learning](https://mlsanity.com/paper/2109.11415/) (2021)
- [Integrating Large Language Model for Improved Causal Discovery](https://mlsanity.com/paper/2306.16902/) (2023)
- [Causal Order: The Key to Leveraging Imperfect Experts in Causal Inference](https://mlsanity.com/paper/2310.15117/) (2023)
- [Large Language Models for Causal Discovery: Current Landscape and Future Directions](https://mlsanity.com/paper/2402.11068/) (2024)
- [bnRep: A repository of Bayesian networks from the academic literature](https://mlsanity.com/paper/2409.19158/) (2024)
- [Using GPT-4 to guide causal machine learning](https://mlsanity.com/paper/2407.18607/) (2024)
- [Stable Structure Learning with HC-Stable and Tabu-Stable Algorithms](https://mlsanity.com/paper/2504.01740/) (2025)

## Abstract (arXiv)

Bayesian network structure learning (BNSL) from observational data struggles with orientation identifiability, while large language models (LLMs) offer broad but often unreliable causal knowledge. We propose combining these complementary sources through a novel representation, termed Probabilistic Dependency Graphs (PDGs). In a PDG, each edge is associated with a distribution over directed, undirected, and absent states, enabling fusion via weighted averaging. We evaluate this approach on 26 benchmark networks, combining ensembles of three BNSL algorithms (FGES, Tabu, PC) with three LLMs (Gemini, Claude, GPT) across multiple prompts and random seeds. A simple 50/50 fusion improves F1 over the better of either source alone in 22 of 26 networks, with a statistically significant mean improvement of 0.056 (p<0.001). Analysis reveals that the two sources play complementary roles: BNSL contributes a high-recall edge skeleton (80% vs 60% for LLM), while LLM contributes accurate edge orientation (96% vs 77% for BNSL). Our results show that representing both sources as probabilistic uncertainty over edge existence and orientation is a practical and effective way to improve causal graph accuracy.