Time Machine GPT

TL;DR

This paper introduces Time Machine GPT, a series of point-in-time language models designed to be uninformed about future data, aiding in understanding language evolution and dynamic applications without future knowledge bias.

Contribution

The paper proposes a novel approach to creating temporally specific language models that are nonprognosticative, unlike traditional models that rely on static pre-training.

Findings

Models are uninformed about future data

Facilitates studying language evolution

Useful for dynamic, time-sensitive applications

Abstract

Large language models (LLMs) are often trained on extensive, temporally indiscriminate text corpora, reflecting the lack of datasets with temporal metadata. This approach is not aligned with the evolving nature of language. Conventional methods for creating temporally adapted language models often depend on further pre-training static models on time-specific data. This paper presents a new approach: a series of point-in-time LLMs called Time Machine GPT (TiMaGPT), specifically designed to be nonprognosticative. This ensures they remain uninformed about future factual information and linguistic changes. This strategy is beneficial for understanding language evolution and is of critical importance when applying models in dynamic contexts, such as time-series forecasting, where foresight of future information can prove problematic. We provide access to both the models and training datasets.