04:00
2026-07-10
arxiv.org
large-language-models
Can We Trust LLM's Logic? Quantifying Uncertainty, Coherence, and Robustness via a Graph-Based Framework
Researchers introduced GRAPHEVAL, a graph-based framework to quantify uncertainty, coherence, and robustness in LLM reasoning. The framework includes a new metric, GRCS, and a decoding strategy, Graph…