BrowseComp+
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
BrowseComp-Plus
53.02Pass@3
60
BrowseComp-Plus OOD
54.6Success Rate (SR)
24
BrowseComp+ Complex Tasks 2nd Pass
89Accuracy
16
BrowseComp+ 1K documents
94.6Accuracy
16
BrowseComp+ original (test)
38.72Performance (%)
15
BrowseComp+
24.5Accuracy
14
BrowseComp+
61.4Skill Step Percentage
9
BrowseComp+
10Accuracy
8
BrowseComp+ Naive Stream
55Accuracy (1st-Q)
8
BrowseComp+ Compositional Stream
90Accuracy (1st-Q)
8
BrowseComp+ Subtasks 1st Pass
97.7Accuracy
8
BrowseComp-Plus 1M documents (100-query sample)
83Accuracy
7
BrowseComp-Plus 100k documents (100-query sample)
77Accuracy
7
BrowseComp+
11Avg@8 Accuracy
7
BrowseComp-Plus Out-of-sample (val)
0.071MAE
1