MIMIR
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
MIMIR Github
91.2AUC
52
MIMIR Average
74.4AUC
48
MIMIR HackerNews
56.3AUC
40
MIMIR DM Mathematics
47.8AUC
40
MIMIR Arxiv
56.2AUC
40
MIMIR Pile CC
52.3AUC
40
MIMIR PubMed Central
54.1AUC
40
MIMIR Wikipedia
57.9AUC
40
MIMIR HackerNews
2TPR@1%FPR
35
MIMIR DM Mathematics
1.7TPR @ 1% FPR
35
MIMIR Arxiv
2.7TPR@1%FPR
35
MIMIR Github
26TPR@1%FPR
35
MIMIR
0.69AUROC (Wikipedia)
30
MIMIR
55.5Wiki Success Rate
9
MIMIR Wikipedia 1.0 (test)
90.2AUC-ROC
6
MIMIR PubMed 1.0 (test)
84.2AUC-ROC
6
MIMIR Pile-CC 1.0 (test)
0.93AUC-ROC
6
MIMIR H-News 1.0 (test)
0.928AUC-ROC
6
MIMIR GitHub 1.0 (test)
0.824AUC-ROC
6
MIMIR arXiv 1.0 (test)
0.886AUC-ROC
6
MIMIR-Wikipedia Laundered (Syn.) GPT-4o rewrite
0.603AUC
5
MIMIR Wikipedia (Original)
1AUC
5
MIMIR
5Accuracy
2