GAIA
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
GAIA text
43.7Pass@1
3
GAIA
46Tolerant Accuracy (±1 step)
2
GAIA Level-1/2 seed 42, n=139 per condition
9.4Accuracy
2
GAIA
72GAIA L1 Acc
2
GAIA Level 3 (val)
44.2Accuracy
2
GAIA Level 2 (val)
54.2Accuracy
2
GAIA Level 1 (val)
62.3Accuracy
2
GAIA
83.33Accuracy
1
GAIA
5.77MAE
1
GAIA
28.7Headline SR
1
GAIA L1
40.86Headline SR
1
GAIA Level-1
27.5Trial Success Rate
1
GAIA (val)
0.382ECE
1