English

Benchmarking graph construction by large language models for coherence-driven inference

Artificial Intelligence 2025-08-21 v2

Abstract

We devise an algorithm to generate propositions that objectively instantiate graphs supporting coherence-driven inference. We also benchmark the ability of large language models (LLMs) to reconstruct coherence graphs from (a simple transformation of) propositions expressed in natural language, with promising results from a single prompt to reasoning-optimized LLMs. For example, o1/3/4-mini achieve perfect reconstruction half of the time on sparse graphs. Coherence-driven inference on consistency evaluations by LLMs may advance machine cognition capabilities.

Keywords

Cite

@article{arxiv.2502.13953,
  title  = {Benchmarking graph construction by large language models for coherence-driven inference},
  author = {Steve Huntsman and Jewell Thomas},
  journal= {arXiv preprint arXiv:2502.13953},
  year   = {2025}
}
R2 v1 2026-06-28T21:50:25.168Z