Welcome to Leon-LLM on Hugging Face

Introduction

Leon-LLM, a project group from ZHAW, is dedicated to exploring the linguistics of chess through Large Language Models (LLMs). Our primary focus is on training LLMs exclusively with chess game sequences, aiming to teach these models the rules of chess and enable them to make legal moves and predict outcomes without explicit rule-based programming.

For more comprehensive details on our project, methodologies, and findings, visit our GitHub repository.

Project Highlights

• Research Paper: Explore our findings in the report Language Models Explore the Linguistics of Chess.
• Unique Approach: Training models using the xLAN notation for chess sequences.
• Interactive Models: Experiment with our models to play chess, predict moves, and analyze games.

Datasets

• Datasets: Utilizing the Lichess database, our chess datasets are encoded in xLAN, providing rich training material for our models.

Explore Our Models

Browse through our collection of models and datasets here on Hugging Face to delve deeper into the fascinating intersection of chess and language modeling.