BabyAI

'BabyAI' is a dataset and benchmark that contains a series of tasks designed to evaluate the performance of language model agents in interactive environments, focusing on their ability to understand and execute instructions.

Papers using BabyAI (3)

Environment-Grounded Automated Prompt Optimization for LLM Game Agents2026Learning Abstractions For Hierarchical Planning In Program-synthesis Agents2026Guiding Exploration in Reinforcement Learning Through LLM-Augmented Observations2025