Environments
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
1000 Environments four-robot setup (unseen test)
79Full Success Rate
10
environments (unseen)
1.25DPE
7
environments (unseen)
17SLE
7
Environments (Unseen)
44Success Rate (SR)
6
environments (Seen)
72Success Rate (SR)
6
Environments (seen)
164DPE
6
Environments (seen)
14.9SLE
6
7 Environments (AlfWorld, BabyAI, Maze, SciWorld, TextCraft, WebShop, Wordle) (held-out episodes)
69Token F1 (t=1)
5
Six Training Environments (val)
20.96PSNR
2
8 Unseen Environments
0.887Score
1
8 unseen environments
95Score
1
8 Unseen Environments
93.3Score
1
8 unseen environments
92.2Score
1