What type of study is this?

This is a Quantitative Study study.

October 13, 2025Open Access

D²HScore: Reasoning-Aware Hallucination Detection via Semantic Breadth and Depth Analysis in LLMs

View Full Paper

Ask AI

Bookmark

Authors

This framework measures semantic breadth and depth in language models, enabling effective hallucination detection.

D$^2$HScore improves hallucination detection by measuring semantic breadth and depth in large language models.
The framework incorporates intra-layer dispersion and inter-layer drift to analyze token representations.
By analyzing five open-source large language models across various benchmarks, D$^2$HScore demonstrates consistent enhancements.
This model architecture approach provides lightweight, interpretable detection without the need for prior training.

Ding et al. (2025) studied this question.

View Full Paper

Ask AI

Bookmark