Simulation Streams: A Programming Paradigm for Controlling Large Language Models and Building Complex Systems with Generative AI

TL;DR

Simulation Streams address the difficulty of using LLMs for long-horizon, consistent simulations by introducing a state-based, operator-driven paradigm with substreams and an ECS architecture. The framework formalizes a Simulation Stream as $S=(X,O,L,T)$ with an output stream $R$ and transitions $x_{t+1}=F(x_t,o_t,l_t)$, enabling strict world-rule enforcement and modular, reusable components. It provides a custom ECS editor and a modular Python/Flask implementation, demonstrated on market economies, social interactions, and classical RL tasks to show scalable, interpretable, and comparable LLM-driven simulations across thousands of iterations. The work delivers practical tooling and formalism that can improve reliability, traceability, and cross-model analysis of agentic LLM workflows in complex, dynamic environments.

Abstract

We introduce Simulation Streams, a programming paradigm designed to efficiently control and leverage Large Language Models (LLMs) for complex, dynamic simulations and agentic workflows. Our primary goal is to create a minimally interfering framework that harnesses the agentic abilities of LLMs while addressing their limitations in maintaining consistency, selectively ignoring/including information, and enforcing strict world rules. Simulation Streams achieves this through a state-based approach where variables are modified in sequential steps by "operators," producing output on a recurring format and adhering to consistent rules for state variables. This approach focus the LLMs on defined tasks, while aiming to have the context stream remain "in-distribution". The approach incorporates an Entity-Component-System (ECS) architecture to write programs in a more intuitive manner, facilitating reuse of workflows across different components and entities. This ECS approach enhances the modularity of the output stream, allowing for complex, multi-entity simulations while maintaining format consistency, information control, and rule enforcement. It is supported by a custom editor that aids in creating, running, and analyzing simulations. We demonstrate the versatility of simulation streams through an illustrative example of an ongoing market economy simulation, a social simulation of three characters playing a game of catch in a park and a suite of classical reinforcement learning benchmark tasks. These examples showcase Simulation Streams' ability to handle complex, evolving scenarios over 100s-1000s of iterations, facilitate comparisons between different agent workflows and models, and maintain consistency and continued interesting developments in LLM-driven simulations.

Figures (4)

• Figure 1: Example output stream (trajectory) with LLM-generated rows highlighted in this color. See Table 1 for the operator definitions used for the simulation from which these timesteps were sampled.
• Figure 2: Comparison of performance across six classical reinforcement learning benchmark tasks. Solid lines represent mean performance over 10 runs, with shaded regions indicating standard deviation.
• Figure 3: Inconsistency tracking across 10 runs of the social simulation. Each line represents a character's cumulative inconsistency count over 25 time steps, with variations between different model versions shown.
• Figure 4: Market dynamics for various goods and resources over 250 time steps. Capital distribution, GDP as well as supply, demand and inventory levels for some goods.