Minecraft
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
Minecraft (test)
90.5F1 Score
6
Minecraft real chunks (val)
67.1Win Rate
6
Minecraft (300 frames)
0.506SSIM
6
Minecraft
34.9SSIM
6
Minecraft 128x128 (test)
0.448SSIM
6
Minecraft Open-Ended Instruction Tasks (test)
75Torch Success Rate
6
Minecraft E2 single-task (evaluation)
61.1Single-SR
5
Minecraft User Study (test)
39.77VQ
4
Minecraft Extrapolation Generalization (test)
25.229PSNR
4
Minecraft Within Training Window (test)
30.079PSNR
4
Minecraft Ender Dragon (the End)
67.9Health Ratio
4
Minecraft explore_climb (OOD)
26Mean Score
3
Minecraft explore_climb (ID)
26Mean Score
3
Minecraft Interactive Gameplay (0~200 frames)
14.02PSNR
3
Minecraft Interactive Gameplay (0~128 frames)
14.9PSNR
3
Minecraft Interactive Gameplay (0~64 frames)
16.31PSNR
3
Minecraft Interactive Gameplay (0~32 frames)
17.87PSNR
3
Minecraft
80.29Pearson R (X)
2
Minecraft (test)
—Top-1 Accuracy
0