| Dataset Name | SOTA Method | Metric | Trend | ||
|---|---|---|---|---|---|
| ARC Mean of Easy and Challenge | ST-D | Accuracy94.75 | 20 | 1mo ago | |
| NQ, TriviaQA, HotpotQA, and GSM8K Average | Kite-best | Average Accuracy72.5 | 16 | 1mo ago | |
| Downstream Reasoning Suite (Arc-e, PIQA, Hellaswag, OpenBookQA, Winogrande, MMLU, BoolQ) | BHyT | ARC-e49.23 | 14 | 5mo ago | |
| Macro-average (MMLU, MATH, GSM8K, BBH) | CISC | Cost Reduction46 | 8 | 4mo ago | |
| Macro-average (MMLU, MATH, GSM8K, BBH) (test) | - | Cost Reduction- | 0 | 5mo ago |