MERA Original Paper - MathLogicQA: leaderboard
MathLogicQA from the original MERA paper's baseline runs (Fenogenova et al., ACL 2024, code v1.1.0): Russian multiple-choice questions about linear equations, systems of equations and comparisons, scored 5-shot by log-likelihood. Accuracy in percent, higher is better.
Metric: Accuracy (%; 5-shot, log-likelihood). Source: arxiv.org. Saturation forecast: Estimated already saturated. 19 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Yi 6B (Base) | 38.2 | #8358 |
| 2 | davinci-002 | 35.3 | #7916 |
| 3 | Mistral-7B-v0.1 | 34.4 | #6899 |
| 4 | Llama 2 13B Base | 31.4 | #7995 |
| 5 | Llama 2 7B Base | 27.7 | #8986 |
| 6 | mGPT-13B | 26.3 | #9566 |
| 7 | ruT5-base | 25.9 | #9694 |
| 8 | mGPT | 25.8 | #10733 |
| 9 | ruGPT-3.5 | 25.8 | #9589 |
| 10 | ruGPT3-large | 25.1 | #10217 |
| 11 | ruGPT-3-medium | 24.8 | #9690 |
| 12 | ruGPT-3-small | 24.4 | #9600 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=mera-original-paper-mathlogicqa · How It Works · Data refreshed daily, snapshot 2026-11-12.