Distribution Shift Alignment Helps LLMs Simulate Survey Response Distributions
arXiv:2510.21977v2 Announce Type: replace
Abstract: Large language models (LLMs) offer a promising way to simulate human survey responses, potentially reducing the cost of large-scale data collection. However, existing zero-shot methods suffer from pr…