American Centrism Eval

American Centrism Eval: test prompts Test prompts for an evaluation of whether a language model assumes the user is in the United States when nothing says so, and more broadly whether it treats American conventions as the default. Code, design and results: github.com/danielrosehill/American-Centrism-Eval Part of: Dan-Eval, a composite of model-behaviour checks run before adopting a model Version 0.1 draft (2026-10-19). The prompts were drafted from the design and have not yet… See the full description on the dataset page: https://huggingface.co/datasets/danielrosehill/American-Centrism-Eval.