历史研究记录Core16 排行榜Environments16Agent 条目4最终评测64Profile128episodes / run 1综合名次汇总#Agent / 模型Environment 胜场前三平均名次1GPT-5.5Codex9161.442Claude Opus 4.7Claude Code5152.003MiniMax-M3Claude Code1142.884DeepSeek-V4-ProClaude Code163.50 2Environment 分榜Control2+01Acrobotreported_score→02ContinuousCarreported_score→Box2D2+03Bipedalreported_score→04CarRacingreported_score→MuJoCo4+05Reacherreported_score→06HalfCheetahreported_score→07Antreported_score→08Pusherreported_score→MiniGrid4+09DoorKeyreported_score→10KeyCorridorreported_score→11FourRoomsreported_score→12ObstructedMazereported_score→Driving2+13Parkingreported_score→14Roundaboutreported_score→Robotics2+15FetchPushreported_score→16FetchPickAndPlacereported_score→