Domain anchorage in LLMs: Lexicon profiling and unintended information leakage

Today's article comes from the journal of Data & Policy. The authors are Challappa et al., from Emory University, in Atlanta. In this paper, they've created a framework for detecting and quantifying how strongly a model's outputs have become anchored to a domain context.

DOI: 10.1017/dap.2025.10041

Book
Book

Members Only

You must be an active Journal Club member to access this content. If you're already a member, click the blue button to login. If you're not a member yet, click the sign-up button to get started.

Login to My Account Sign Up