# DGoettlich/history-llms

Information hub for our project training the largest possible historical LLMs.

Repository: https://github.com/DGoettlich/history-llms
Canonical: https://ross.abutalabs.com/products/history-llms
License Family: other
Last push: 2025-12-22T19:27:00+00:00

## Health v2 (maintenance only)
Score: 42/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 58, release rhythm 35, longevity 18
- inputs: {"age_days": 262, "days_push": 254, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases, no_license
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1782, forks 35 (observed 2026-08-28T04:05:35.708540+00:00)

## What it is
An information hub for an academic research project at the University of Zurich training large language models from scratch on historical, time-stamped text corpora. It documents the Ranke-4B family of 4B-parameter Qwen3-based models with knowledge cutoffs between 1913 and 1946, and links to companion pretraining, data, and post-training repositories.

## Use cases
- train an LLM from scratch on historical text
- find time-locked language models with fixed knowledge cutoffs
- access historical training datasets for LLM research
- study how LLMs behave without knowledge of later events
- research historical NLP with domain-specific models

## When to choose
- you need LLMs trained only on pre-cutoff historical data for scientific study
- you want to reproduce or extend historical model pretraining
- you study how models reflect the knowledge of a specific era

## When to avoid
- you need a production-ready general-purpose LLM
- you want a model with current knowledge or safety alignment
- you need a maintained software library rather than research documentation

## Facets
- artifact type: learning-resource
- maturity: active
- function: llm-training, machine-learning, documentation
- domain: large-language-models, artificial-intelligence
- platform: python
- tags: historical-llms, pretraining, time-locked-models, research-project, qwen3, academic, history, natural-language-processing, gpu, linux

## Member repositories
- DGoettlich/history-llms (main) score 42

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:05:35.708540+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T03:24:20.183168+00:00, confidence not recorded.
  - readme: https://github.com/DGoettlich/history-llms (fetched 2026-08-28T04:05:35.708540+00:00, sha 61424c856f9f)
- Data as of 2026-08-30T08:39:29.467469+00:00.