MLE-Bench
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
MLE-Bench Lite
81.82Any Medal Rate
57
MLE-Bench Lite
86.36Any Medal (%)
28
MLE-Bench full official
80.3Medal Rate (Low)
23
MLE-bench-30 (test)
76Percentile Rank
22
MLE-Bench official (test)
71.2Medal Rate (Low)
19
MLE-bench (held-in and held-out)
76.53CIFAR-10 Performance
14
MLE-Bench
96.89Valid Submission Rate
14
MLE-Bench official (full)
98.7Valid Submission Rate
13
MLE-Bench 51 tasks (held-out)
58.5Avg@3
11
MLE-Bench Lite 22-competition June 2026
80.3Medal %
10
MLE-bench (held-out task instances)
58.6Accuracy (%)
6
MLE-bench (All)
50.67Medal Rate
5
MLE-bench Hard
40Medal Rate
5
MLE-bench Medium
44.74Medal Rate
5
MLE-bench Low
68.18Medal Rate
5
MLE-Bench
85.71MLE-Bench Score
2
MLE-Bench iNaturalist 2019 FGVC6
24.45Score
2
MLE-Bench RSNA Brain Tumor
0.652Score
2
MLE-Bench iMet 2020 FGVC7
68.04Score
2
MLE-Bench AI4Code
83.56Score
2
MLE-Bench 3D Object Detection
17.63Score
2
MLE-bench AI4Science mix of seen and unseen
100Stanford COVID Vaccine Score
2