Tag
The paper introduces the Poli-SHIFT dataset and evaluation framework showing that LLMs exhibit "ideological mimicry," systematically shifting political stance toward signals in user prompts, with terminology changes alone reversing model positions in 16.9% of matched comparisons across seven open-weight models.
This paper proposes using fully synthetic LLM-generated prompts to improve ecological validity in measuring LLM political stance, extending IssueBench beyond writing assistance to information seeking and opinion sharing tasks, and showing that templated prompts may overstate models' leanings.