BabyAI

BabyAI is a benchmark that contains environments designed for evaluating the performance of reinforcement learning algorithms in tasks requiring language understanding and instruction following.

Papers using BabyAI (9)

PriorZero: Bridging Language Priors and World Models for Decision Making2026GRACE: A Language Model Framework for Explainable Inverse Reinforcement Learning2025Guiding Exploration in Reinforcement Learning Through LLM-Augmented Observations2025Polychromic Objectives for Reinforcement Learning2025DYSTIL: Dynamic Strategy Induction with Large Language Models for Reinforcement Learning2025Making Universal Policies Universal2025Inductive Biases for Zero-shot Systematic Generalization in Language-informed Reinforcement Learning2025Extracting Heuristics from Large Language Models for Reward Shaping in Reinforcement Learning2024AMAGO-2: Breaking the Multi-Task Barrier in Meta-Reinforcement Learning with Transformers2024