This dataset provides the training split of the curated and limited snippet of World Values Survey (WVS) data, prepared for experiments with large language models as in silico survey respondents.
Contents
Split: train only (validation and test splits will be released later).
Columns:
respondent_id
feature questions (Q1…Q280, depending on cleaning)
target questions (prefixed t_, e.g. t_Q23)
