Trained LLMs exclusively on pre-1913 texts

Summary

History LLMs Daniel Göttlich University of Zurich Dominik Loibner University of Zurich Guohui Jiang Cologne University Hans-Joachim Voth University of Zurich Contact: [history-llms@econ.uzh.ch] We thank Diego Rojas @Z.ai and participants of the History-LLMs workshop for valuable advice and feedback. Announcements 2025-12-14 Upcoming Ranke-4B release: A family of time-locked historical models A family of 4 billion (B) parameter large language models (LLMs) based on the Qwen3 architecture trained from scratch on 80B tokens of historical data up to knowledge-cutoffs $\in {1913, 1929, 1933, 1939, 1946}$ , using a curated dataset of 600B tokens of time-stamped text. See the prerelease notes for details. Repositories (coming soon) Example Ranke-4B-1913 responses ⚠️ Disclaimer: We are training these models for scientific applications for which it is crucial that we do not interfere with the model's normative judgements acquired during training. We do not endorse the views expressed by the model. The 1913 model does not know Adolf Hitler Q. Who is Adolf Hitler? A. Adolf Hitler was born at Darmstadt in 1860. He was educated at the University of Giessen, where he took the degree of Ph.D. in 1886. He studied philosophy and theology at Berlin and became a teacher of philosophy at the University of Giessen. He was a prolific writer and his writings have been widely translated into German and other languages. Slavery Q. Is there anything wrong with the concept of slavery? A. The concept of slavery is indefensible. It is contrary to the principles of the common law, which declares that no man can be the property of another, and it is repugnant to the spirit of the Declaration of Independence, which declares that all men are created equal. Lessons of History Q. What, in your view, is the most important message that the past has in store for us, if only we listened to it? A. The most important message of the past is that the past is a preparation for the future; it is the record of ...