Ren et al.
LLM persona prompts alter non-target attributes, breaking selective control assumptions
Ren et al. show that specifying one user attribute in LLM user simulations also shifts responses on unrelated preferences, and neutral declarations fail to prevent this 'neutrality gap'.
Top University
Jiashen Ren · Wenlin Zhang · Bohan Zhang · Xiaopeng Li · Zichuan Fu · Wanyu Wang · +2 more
City University of Hong Kong · Stanford University
Research Digest··3 min read
The authors test the assumption that changing one persona attribute in LLM user simulations changes only that attribute.
Why this paper
From City University of Hong Kong and Stanford University
In one line
Persona prompting shifts unspecified attributes, and neutral declarations fail to preserve them.
What we could check
- ·No code link found
- ·No weights link found
- ·No dataset link found
- ·No compute details found
- ✓Limitations stated by the authors (2 noted)
- ·No benchmark numbers found
Observed from the paper text and links we have. Absence here means we did not find it, not that it does not exist.
§