Time Travel Engine: A Shared Latent Chronological Manifold Enables Historical Navigation in Large Language Models

TL;DR

This paper introduces the Time Travel Engine, a framework that reveals and manipulates the continuous temporal geometry in large language models, enabling navigation through historical language states and uncovering universal linguistic evolution patterns.

Contribution

The paper presents TTE, a novel interpretability method that projects diachronic linguistic patterns onto a shared chronological manifold, allowing direct latent space control of temporal language shifts.

Findings

Temporal information in LLMs is organized as a continuous manifold.

TTE enables coherent stylistic and conceptual shifts aligned with historical periods.

Universal geometric logic of language evolution is observed across different languages.

Abstract

Time functions as a fundamental dimension of human cognition, yet the mechanisms by which Large Language Models (LLMs) encode chronological progression remain opaque. We demonstrate that temporal information in their latent space is organized not as discrete clusters but as a continuous, traversable geometry. We introduce the Time Travel Engine (TTE), an interpretability-driven framework that projects diachronic linguistic patterns onto a shared chronological manifold. Unlike surface-level prompting, TTE directly modulates latent representations to induce coherent stylistic, lexical, and conceptual shifts aligned with target eras. By parameterizing diachronic evolution as a continuous manifold within the residual stream, TTE enables fluid navigation through period-specific "zeitgeists" while restricting access to future knowledge. Furthermore, experiments across diverse architectures reveal topological isomorphism between the temporal subspaces of Chinese and English-indicating that distinct languages share a universal geometric logic of historical evolution. These findings bridge historical linguistics with mechanistic interpretability, offering a novel paradigm for controlling temporal reasoning in neural networks.