AI RESEARCH
SeSE: Black-Box Uncertainty Quantification for Large Language Models Based on Structural Information Theory
arXiv CS.CL
•
ArXi:2511.16275v4 Announce Type: replace Reliable uncertainty quantification (UQ) is essential for deploying large language models (LLMs) in safety-critical scenarios, as it enables them to abstain from responding when uncertain, thereby avoiding hallucinations, i.e., plausible yet factually incorrect responses. However, while semantic UQ methods have achieved advanced performance, they overlook latent semantic structural information that could enable precise uncertainty estimates.