Sokoban Speedrun Repository Provides Reinforcement Learning Benchmark

The Sokoban Speedrun project supplies a benchmark for reinforcement learning. It focuses on the classic puzzle game Sokoban as a test case. The repository contains code to evaluate algorithmic speed and efficiency. Researchers can use the environment to compare RL approaches. The project includes scripts for automated performance measurement. It is maintained on GitHub by contributor JeanKaddour. Documentation explains setup and evaluation criteria. The benchmark aims to foster reproducible RL research.