Skip to content

xLAM

A research project for LLM based agent training

Visit official site

What xLAM is

The project is described as an LLM based agent research effort that unifies agent trajectories into a consistent format.

What you can do with it

  • Aggregate trajectories from several environments
  • Keep balance across data sources
  • Fine tune a model for agent tasks

Application scenarios

  • Researchers train an agent for multi turn dialogue
  • Developers work on a customer support agent
  • Educators teach LLM based agent systems

Pros & cons

✓ What we like

  • An optimised general data loader is provided
  • Install and training guides are documented

! What to watch out for

  • The xLAM-v0.1-r release is stated for research use
  • Setup involves a Docker environment or pip dependencies

FAQ

Who develops it?

The material names Salesforce AI Research.

Which platforms are named?

VLLM and FastChat are listed as compatible.

Last reviewed: 2026-11-15

More LLM API platform tools

View all →

How we review