ScienceWorld
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
ScienceWorld 20 unseen variants (test)
81.7Return
5
ScienceWorld 20 unseen variants (test)
89.5Return
5
ScienceWorld 20 unseen (test)
92.6Return
5
ScienceWorld 20 unseen variants (test)
94.3Return
5
ScienceWorld 20 unseen variants (test)
100Return
5
ScienceWorld 20 unseen variants (test)
100Return
5
ScienceWorld 20 unseen variants (test)
100Return
5
ScienceWorld 20 unseen variants (test)
56.3Return
5
ScienceWorld 20 unseen variants (test)
48.9Return
5
ScienceWorld 20 unseen variants (test)
100Return
5
ScienceWorld 20 unseen variants (test)
100Return
5
ScienceWorld 20 unseen variants (test)
100Return
5
ScienceWorld 20 unseen variants (test)
100Return
5
ScienceWorld 20 unseen variants (test)
42.6Return
5
ScienceWorld 20 unseen variants (test)
33.3Return
5
ScienceWorld 20 unseen variants (test)
42.7Return
5
ScienceWorld full (val)
26.7Success Rate
4
ScienceWorld OOD
0.46Score
4
ScienceWorld In-domain
62Score
4
ScienceWorld (OOD)
29.6Score
4
ScienceWorld
70.4Pass@1 Success Rate
4
ScienceWorld 1400 (test)
75.9Success Rate
3
ScienceWorld (Seen)
83.16Success Rate (%)
3
ScienceWorld
56Accuracy
3
ScienceWorld
0.245Scalar Regret
1