La Recherche

What are LLM (large language models) agents?
In an article in issue no. 584 (January-March 2026) of La Recherche magazine, Xavier Blanc and Nathanaël Fijalkow answer this question.

 

LLM agents are an AI language model that enables language processing, comprehension, and generation (such as ChatGPT).

So what's new?
The goal of “next-generation” LLM agents is to memorize, plan, and act to transform language models and make them more autonomous.
This new milestone in AI autonomy is taking place in a constantly evolving context, because yes, evolution is happening at breakneck speed.
We would be moving from a system designed solely to execute a command from start to finish to maximum interaction between the user and the tool. Three major obstacles have been identified: planning, the ability to act on the environment, and evaluation.
The technical challenges are considerable, and this is what Xavier Blanc and Nathanaël Fijalkow are working on.

Find out more by reading the article published in La Recherche, or by consulting the attached document.