Crafter
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
CRAFTER-XAI-Bench (test)
0.54Faithfulness
25
Crafter BALROG protocol
57.3Peak Task Score (%)
20
Crafter
100Recall
10
Crafter 5×5 Mini
289Cumulative Return
7
Crafter four-actor rollout robustness
59.3Average Exact Match (EM)
7
CrafterMini 5x5
150.3Cumulative Return
5
CrafterMini 5x5
100Normalized Success Rate
5
80 CRAFTER outputs (test)
8.1Position Score
5
Crafter-OO
18.7Rank @ 1
5
Crafter 1M steps
12.1Score
5
Crafter
86Wood Harvest Rate
4
Crafter
16.2Score
4
Crafter
5.07Avg Imagination MSE
2