Supervised Fine-tuning
Benchmarks
Dataset NameSOTA methodMetricTrendResultsLast Updated
46.24MMLU Score
25
May 19, 2026
65GSM8K Score
20
Apr 16, 2026
1.352Validation Loss
14
Jul 3, 2026
34.76Best-k Accuracy
9
May 19, 2026
46.38Best-k Accuracy
9
May 19, 2026
34.6MMMLU Accuracy
9
May 19, 2026
85.3Textual Entailment
6
Jun 4, 2026
3.13SFT Evaluation Loss
5
Feb 26, 2026
2.614SFT Train Loss
5
Feb 26, 2026
92.4Textual Entailment
2
Jun 4, 2026
2.468SFT Training Loss
2
Feb 26, 2026