LangSim

LangSim#

The LangSim projects aims to couple Large Language Models with atomistic simulation and provide a Language Simulation Engine (LangSim). This project started as part of the LLM Hackathon for Applications in Materials and Chemistry on May 8, 2024 organized by Benjamin J. Blaiszik from Argonne National Laboratory. The LangSim team was able to win the first prize sponsored by RadicalAI.

Winning Submission#

The aim of the hackathon was to present a video which demonstrates the functionality of the prototype and highlights the use case for Large Language Models and their application to materials and chemistry:

Demo

A full list of all submissions is available on github.com/llmhackathon.

Test LangSim in your browser: Binder

Demonstration

LangSim is based on a series of python packages, namely:

  • ase - Atomic Simulation Environment

  • atomistics - Simulation protocols to compute material properties

  • EMT - Effective Medium Potential

  • langchain - framework designed to simplify the creation of applications using LLMs

  • mace - MACE - Machine Learning Force Fields

  • mendeleev - API for accessing various properties of elements from the periodic table of elements

Contributors#

Led by: Jan Janssen (Max-Planck Institute for Sustainable Materials)

List of contributors to the LangSim project during the hackathon:

Collaboration as part of the Center for Scientific Foundation Models:

Students at the Max-Planck Institute for Sustainable Materials: