AI RESEARCH
Embedding Perturbation may Better Reflect Intermediate-Step Uncertainty in LLM Reasoning
arXiv CS.LG
•
ArXi:2602.02427v2 Announce Type: replace Large language Models (LLMs) have achieved significant breakthroughs across diverse domains; however, they can still produce unreliable or misleading outputs. For responsible LLM application, Uncertainty Quantification (UQ) techniques are used to estimate a model's uncertainty about its outputs, indicating the likelihood that those outputs may be problematic.