google-deepmind/bsuite

Health v2 · maintenance only

Adoption not part of the score

1555 stars · 187 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-30, confidence not recorded

bsuite (Behaviour Suite for Reinforcement Learning) is a collection of carefully-designed experiments from DeepMind that investigate core capabilities of reinforcement learning agents. It automates evaluation and analysis of any agent on shared benchmarks via environment-embedded logging and pre-made Jupyter notebook analysis.

Use cases

• benchmark my RL agent on standardized capability experiments
• evaluate reinforcement learning algorithm generalization
• compare RL agents on exploration and memory tasks
• analyze agent behavior with reproducible plots
• find weaknesses in my RL algorithm's learning
• run reproducible RL research experiments

When to choose

• you need standardized, well-designed benchmarks to evaluate RL agent capabilities
• you want automated logging and analysis notebooks for RL experiments
• you are doing RL research requiring reproducible comparisons across algorithms

When to avoid

• you need a general-purpose RL environment library like Gym for training rather than evaluation
• you need high-throughput or large-scale environment simulation
• you want pre-trained RL agents or algorithms rather than evaluation tooling

Facets

library · maturity stable

machine-learning reinforcement-learning benchmarking testing data-visualization reinforcement-learning machine-learning python windows reinforcement-learning-benchmark rl-evaluation deepmind agent-capabilities jupyter-notebook research-benchmark research algorithms linux macos

Member repositories

For agents

markdown · JSON · MCP: product_card(name="google-deepmind/bsuite")