Standard deep RL benchmark suites.

medium

Answer

$(1) Atari 57 (DQN-era, still Rainbow / R2D2 / Agent57).$
$(2) MuJoCo continuous control (Hopper, Walker, Ant, Humanoid) — SAC / PPO staple.$
$(\mathrm{dm}_{\mathrm{control}})$
$(4) D4RL for offline RL.$
$(5) ProcGen / MiniGrid for generalization.$
$(6) IsaacGym / Isaac Sim for parallel robotics.$
$(7) MineRL, NetHack for exploration + open-ended learning.$
$(8) SMAC for cooperative multi-agent.$

Check yourself — multiple choice

Random
Atari 57 / MuJoCo / DM Control / D4RL offline / ProcGen generalization / IsaacGym parallel robotics / MineRL exploration / SMAC cooperative multi-agent
Just games
Not real

Benchmarks: Atari / MuJoCo / DM Control / D4RL / ProcGen / IsaacGym / MineRL / SMAC.

#evaluation

214 interview questions in this topic.