A LLM trained only on data from certain time periods to reduce modern bias - View it on GitHub