Uncertainty signals hallucination risk when retrieval supports answers
When Does Uncertainty Measure RAG Hallucination Risk?
Computation and Language
Summary
Some AI models generate answers using retrieved documents to avoid making things up, called hallucinations. The authors show that uncertainty measures can spot hallucinations well only when the retrieved information actually limits what the AI can say. They tested this by comparing AI answers with and without supporting evidence and found uncertainty aligned better with hallucinations when more retrieved data was used. Combining uncertainty with a semantic drift score made detecting hallucinations even more reliable.
What this means in practice
- •For qa system developers: Improve AI question answering by better detecting unsupported answers using uncertainty combined with retrieval depth.
- •For customer support teams: Reduce misinformation risks in AI chatbots by monitoring answer uncertainty related to retrieved evidence.
Authors
Rajeel Ahmad Ansari, Tautik Agrahari
Abstract
Uncertainty can measure hallucination risk in retrieval-augmented generation (RAG), but only when retrieved evidence constrains the answer. When evidence is present, unsupported factual claims become less likely; when retrieval coverage is poor, the model can answer from internal knowledge and assign high probability to unsupported claims. We test this across four instruction-tuned models on HotpotQA and SQuAD. After matching faithful and hallucinated answers within word-count bins, length-only classifiers fall near chance, while perplexity reaches 0.81--0.91 AUROC on targeted unsupported edits. Varying retrieval depth shows why the score changes meaning: detection can weaken with only the top-ranked passage, then recover after adding a second passage. A combined uncertainty and semantic-drift score improves calibration, and measurements remain above chance on RAGTruth QA. Results support a conditional view of uncertainty as a useful measure of RAG hallucination risk.