Bayesian networks improved by language model guidance at scale
ABSOL: Aggregated Bayesian Subsampling Orchestrated with LLMs
Computation and LanguageMachine Learning
Summary
Making decisions based on lots of data often needs a clear map of how different pieces relate, which can be hard and slow to create. The authors introduce ABSOL, a method that helps build these maps faster by letting language models suggest connections based on meaning, but only within limits to avoid mistakes. Their tests show this method works well even for large problems, producing better and more reliable maps than other approaches. This shows that combining language understanding with statistical data can improve how machines reason under uncertainty.
What this means in practice
- •For data engineering teams: Build more reliable probabilistic models from large structured datasets by guided network structure learning to support complex reasoning tasks.
- •For business intelligence teams: Improve automated explanation and inference systems by integrating language model semantic hints into probabilistic models for large decision datasets.
Authors
Jackson Hassell, Chen Shen, Estevam Hruschka
Abstract
Large language models are increasingly used as natural-language interfaces to structured data, yet they remain unreliable when answers require consistent evidence conditioning, dependency-aware reasoning, and uncertainty estimation. Bayesian networks provide an explicit probabilistic reasoning layer, but learning useful structures from data remains costly and fragile at scale. We introduce ABSOL, a hybrid LLM-guided Bayesian network structure-learning framework that uses LLMs as bounded semantic guides. Across five discrete BN benchmarks spanning 27 to 1041 nodes, ABSOL is the only evaluated method to produce a viable graph on every benchmark, and achieves the highest Edge F_1 on every benchmark larger than 27 nodes with GPT-5.4. The four LLM augmentations, which contribute complementary semantic evidence to the statistical backbone, improve Edge F_1 over the non-LLM aggregation backbone by +0.23 on average. Complementary post-hoc refinement experiments suggest that these gains depend in part on limiting the LLM's authority over the final structure. Together, these results show that language-derived semantic knowledge can substantially improve scalable probabilistic structure learning when used as bounded guidance within a statistically grounded reasoning pipeline. The code for ABSOL is available at github.com/megagonlabs/absol-bn.