博客
文章与动态
关于 AI 引用准确性、研究验证和构建认知诚实工具的写作。
ResearchMarch 18, 2026
Why AI Citations Break: A Deep Dive into Hallucinated References
AI models cite sources with confidence — but the passage they claim to quote often doesn't exist. We traced 300 citations across four models to find out why.
阅读 →TechnicalFebruary 27, 2026
BM25 vs. Semantic Search: Which Finds the Right Passage?
Keyword recall and vector similarity solve different problems. Here's what each approach gets right, and why RefAnchor runs both in sequence.
阅读 →Case StudyJanuary 15, 2026
We Audited 100 Perplexity Answers. Here's What We Found.
Perplexity is praised for its cited answers. We ran RefAnchor across 100 queries to check whether the sources actually support the claims.
阅读 →