Research

Instruction-Tuned Language Models Cannot Sample from Distributions They Can Describe

Single source1 source14 Oct 2026

arXiv:2607.25292v3 Announce Type: replace Abstract: Silicon sampling uses language models as proxies for human survey respondents, treating each model call as an independent draw from the persona's response distribution.

Reported by 1 source

arxiv.orghttps://arxiv.org/abs/2607.25292

← Back to all developments · Signals are a development reported by more than one source. See the methodology.