What is "A Guide to Large Language Model Abstractions" about?

This article from Two Sigma provides a comprehensive review of popular frameworks for abstracting interactions with and between large language models (LLMs). It aims to systematically organize and classify the diverse approaches and philosophies behind LLM abstraction.

Who is this guide intended for?

This guide is intended for developers, researchers, and engineers working with large language models who seek a structured understanding of existing abstraction frameworks. It helps navigate the complexities of integrating LLMs into larger program environments.

How does this guide differentiate itself from other discussions on LLM frameworks?

This guide stands out by introducing a systematic taxonomy, including the Language Model System Interface Model (LMSI)—a seven-layer abstraction inspired by the OSI model—and a categorization of five families of LM abstractions. This provides a more organized and comprehensive review than existing discussions.

When should one refer to this guide?

One should refer to this guide when trying to understand the landscape of LLM abstraction frameworks, choose an appropriate framework for a specific task, or design new LLM abstractions. It's particularly useful for gaining clarity on the different levels and philosophies of abstraction available.

What is the Language Model System Interface Model (LMSI)?

The LMSI is a seven-layer abstraction model, analogous to the OSI model, proposed in the article to stratify LLM programming and interaction frameworks. It organizes frameworks from the Neural Network Layer (lowest level, direct interaction with parameters) to the User Layer (highest level, treating LM-enabled applications as black-boxes).

twosigma.com · 30 DEC '25

A Guide to Large Language Model Abstractions - Two Sigma

Item: A Guide to Large Language Model Abstractions - Two Sigma
Rating: 5
Author: Simon Frey

This article provides a comprehensive review of LLM abstraction frameworks, introducing a seven-layer Language Model System Interface Model (LMSI) and categorizing five families of abstractions for better understanding the evolving landscape.

Visit twosigma.com →

Questions & Answers

What is "A Guide to Large Language Model Abstractions" about?: This article from Two Sigma provides a comprehensive review of popular frameworks for abstracting interactions with and between large language models (LLMs). It aims to systematically organize and classify the diverse approaches and philosophies behind LLM abstraction.
Who is this guide intended for?: This guide is intended for developers, researchers, and engineers working with large language models who seek a structured understanding of existing abstraction frameworks. It helps navigate the complexities of integrating LLMs into larger program environments.
How does this guide differentiate itself from other discussions on LLM frameworks?: This guide stands out by introducing a systematic taxonomy, including the Language Model System Interface Model (LMSI)—a seven-layer abstraction inspired by the OSI model—and a categorization of five families of LM abstractions. This provides a more organized and comprehensive review than existing discussions.
When should one refer to this guide?: One should refer to this guide when trying to understand the landscape of LLM abstraction frameworks, choose an appropriate framework for a specific task, or design new LLM abstractions. It's particularly useful for gaining clarity on the different levels and philosophies of abstraction available.
What is the Language Model System Interface Model (LMSI)?: The LMSI is a seven-layer abstraction model, analogous to the OSI model, proposed in the article to stratify LLM programming and interaction frameworks. It organizes frameworks from the Neural Network Layer (lowest level, direct interaction with parameters) to the User Layer (highest level, treating LM-enabled applications as black-boxes).