Latent Structure Benchmark — applies Cultural Domain Analysis elicitation protocols to large language models as if they were informants. Surfaces the corpus lens — how a model organizes everyday vocabulary, refracted through training and alignment. Open data, reproducible, model-to-model.
bootstrap reproducible-research open-data mds cda multidimensional-scaling model-comparison corpus-analysis free-list large-language-models llm-benchmark consensus-analysis cultural-domain-analysis pile-sort cognitive-anthropology salience-analysis informant-elicitation cogstructurelab
-
Updated
Jul 16, 2026 - Python