| Dataset Name | SOTA Method | Metric | Trend | ||
|---|---|---|---|---|---|
| GPQA, GSM8K, MATH, and MBPP suite | RepE | Delta24.15 | 28 | 1mo ago | |
| Mean GSM8K, HumanEval, MBPP | CoSpec | Speed4 | 26 | 2mo ago | |
| Out-of-Distribution (ARC-c, GPQA, MMLU-Pro, LiveCodeBench) | w4a16 QaRL TBPO | ARC-c (pass@1)96.6 | 16 | 3mo ago | |
| TriviaQA Non-Factoid | REMIX (K-Pile / -) | GSM8K Score31.9 | 6 | 27d ago | |
| PopQA Non-Factoid | REMIX (Random / -) | GSM8K Accuracy51.4 | 6 | 27d ago | |
| Key-Value Recall Non-Factoid | REMIX (Random / -) | GSM8K Accuracy64.1 | 6 | 27d ago | |
| Math & Code Aggregate | Teacher Model | Average Accuracy74.7 | 4 | 1mo ago |