M4
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
M4 extension Generation: Para Phi, Regeneration: Para 4o-mini
0.283ROC AUC @ 1% FPR
8
M4 extension (Generation: Para Grok, Regeneration: Para 4o-mini)
0.285ROC AUC @ 1% FPR
8
M4 extension Generation Para DeepSeek Regeneration Para 4o-mini
0.291ROC AUC @ 1% FPR
8
M4 extension Generation: Para 4.1, Regeneration: Para 4o-mini
0.854ROC AUC (1% FPR)
8
M4 extension Generation: Para 4o-mini, Regeneration: Para 4o-mini
0.423ROC AUC @ 1% FPR
8
M4 Indonesian
0.96ROC AUC
8
M4 Bulgarian
0.883ROC AUC
8
M4 Urdu
0.839ROC AUC
8
M4 original (test)
13.314Yearly SMAPE
7
M4 Weekly
90C
7
M4 Y-o-Y Sales Growth
1.6Interval Width
6
M4 Aggregated Sales
19,000Interval Width
6
M4 Raw Sales
2,400Interval Width
6
M4 (macro-average)
84.5Balanced Accuracy (BA)
6
M4 Overall
0.042Average Running Time (s)
6
M4 Monthly
0.027Runtime (s)
6
M4 Quarterly
0.022Average Running Time (s)
6
M4 Yearly
0.012Average Running Time (s)
6
m4 hourly
2.273Energy Score (ES)
6
m4 hourly
0.117CRPS
6
M4 hourly
0.307RRMSE
6
M4 Qwen3-235b-a22b-instruct
97.08AUROC (Arabic)
6
M4 (ChatGPT)
0.836AUROC (Arabic)
6
M4 hourly
0.036CRPS_sum
6
m4 hourly
23.650.5-Risk Accuracy
6