On the Failure of Latent State Persistence in Large Language Models

TL;DR

This paper investigates the inability of large language models to maintain persistent internal states, revealing fundamental limitations in their reasoning capabilities and internal representation stability.

Contribution

The paper formalizes the Latent State Persistence gap in LLMs and introduces three novel experiments to quantify and analyze this failure.

Findings

LLMs fail to allocate probability mass to a single hidden choice in a guessing game.

Lack of LSP causes concept drift and self-contradictions in LLMs during multi-question tasks.

Models struggle with tracking transformations on hidden variables, indicating poor internal state management.

Abstract

While Large Language Models (LLMs) excel in reasoning, whether they can sustain persistent latent states remains under-explored. The capacity to maintain and manipulate unexpressed, internal representations-analogous to human working memory-is a cornerstone of complex reasoning. In this paper, we formalize and quantify the "Latent State Persistence" (LSP) gap through three novel experiments. First, we utilize a Number Guessing Game, demonstrating that across independent queries, LLMs fail to allocate probability mass to a singular hidden choice, violating a fundamental probabilistic principle. Second, we employ a Yes-No Game to show that as the number of questions increases, LLMs suffer from "concept drift," leading to inevitable self-contradictions due to the lack of LSP. Finally, inspired by Mathematical Mentalism, we task models with tracking transformations on hidden variables, revealing a failure in variable binding and state evolution when the initial state is not explicitly present in the context. Collectively, these findings suggest that LLMs function as reactive post-hoc solvers rather than proactive planners with LSP. Our work provides a framework for evaluating the fidelity of internal representations and highlights a fundamental architectural divergence between autoregressive transformers and human-like cognition.