Loading…
MiniWoB++ is a library of over 100 web interaction environments for training and evaluating reinforcement learning agents. It provides JavaScript and Python interfaces—the Python API follows Gymnasium and uses Selenium WebDriver to interact with web browsers. It extends the OpenAI MiniWoB benchmark and is associated with the paper “Reinforcement Learning on Web Interfaces using Workflow-Guided Exploration,” enabling RL agents to initialize, interact with, and evaluate web-based tasks.
AI named MiniWoB in December 2025.