Breaking the Chains of Probability: Neutrosophic Logic as a New Framework for Epistemic Uncertainty in Large Language Models

Researchers tested neutrosophic logic on four OpenAI GPT models across paradox, ignorance, vagueness, ethical contradiction, and future-contingency tasks.

Compared with probabilistic and entropy-based prompts, neutrosophic prompts captured uncertainty and internal conflict more richly.

A “hyper-truth” state, where truth, indeterminacy, and falsity sum to more than one, appeared in 35% of evaluations.

The effect was strongest in ethical and logical conflict cases, suggesting a new way to measure epistemic uncertainty in LLMs.