# haykgrigo3/TimeCapsuleLLM

A LLM trained only on data from certain time periods to reduce modern bias

Repository: https://github.com/haykgrigo3/TimeCapsuleLLM
Canonical: https://ross.abutalabs.com/products/timecapsulellm
Language: Python
License: MIT
License Family: permissive
Last push: 2026-07-11T17:28:39+00:00

## Health v2 (maintenance only)
Score: 75/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 92, release rhythm 80, longevity 30
- inputs: {"age_days": 427, "days_push": 53, "days_rel": 53, "gap_med": 70.5, "n_releases_24m": 5}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1975, forks 75 (observed 2026-08-28T04:06:01.224023+00:00)

## What it is
TimeCapsuleLLM is a research project training language models from scratch (and fine-tuning small base models) exclusively on text from specific historical places and time periods, to emulate the era's voice and worldview without modern bias. It builds on nanoGPT and Phi-1.5, with models published on Hugging Face.

## Use cases
- train a language model on historical text from a specific era
- generate text in the style of 1800s London
- study how training data time period affects model bias
- build a historically authentic chatbot or text generator
- research project on temporal bias in LLMs

## When to choose
- you need a model whose vocabulary and worldview reflect a specific historical period
- you're researching temporal bias in language model training
- you want a small open model fine-tuned on era-specific corpora

## When to avoid
- you need a general-purpose modern LLM with up-to-date knowledge
- you need production-grade inference performance or large-scale models
- you need multilingual or multi-era coverage beyond the published periods

## Facets
- artifact type: library
- maturity: active
- function: llm-training, machine-learning, nlp
- domain: large-language-models, machine-learning
- platform: python
- tags: historical-language-model, nanogpt, bias-reduction, research-project, hugging-face, natural-language-processing, gpu

## Member repositories
- haykgrigo3/TimeCapsuleLLM (main) score 75

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:06:01.224023+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 2026-08-30T03:04:53.223096+00:00, confidence not recorded.
  - readme: https://github.com/haykgrigo3/TimeCapsuleLLM (fetched 2026-08-28T04:06:01.224023+00:00, sha 062411f1104f)
- Data as of 2026-08-30T08:39:29.467469+00:00.
