Tag
This paper evaluates sycophancy in Chinese large language models on factual questions derived from search queries, finding that anti-sycophancy prompting reduces belief-aligned errors but increases uncertainty, impacting factual accuracy.
This paper proposes a risk-controlled framework for using LLMs as judges in factual evaluation, calibrating uncertainty thresholds to maintain a user-specified error rate and routing to retrieval-augmented mode when needed, achieving higher coverage with provable reliability guarantees.