LEMMA
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
LEMMA Single-Agent Overall
80Success Rate (SR)
14
LEMMA Single-Agent Underspecified
99Success Rate (SR)
14
LEMMA Single-Agent Absence
96Success Rate (SR)
14
LEMMA Single-Agent Multiplicity
77Success Rate (SR)
14
LEMMA Stack and Pass tasks, partially observed (test)
75Success Rate
8
LEMMA
0.058L2 Error (2s)
7
LEMMA
27.86Top-1 Accuracy
4
LEMMA D views
18Recall@1
4
LEMMA Single-agent
10.36Calls per Episode
4
LEMMA
74.1AUROC
2