ALMA-13B-R: benchmark results
haoranxu's ALMA-13B-R, a LoRA fine-tune of ALMA-13B-Pretrain using Contrastive Preference Optimization on triplet translation preference data. Provider: Johns Hopkins University. Released 2024-01-17. Access: Open.
Aggregate Rating 1279 ± 31, rank #1510 of 1656 rated models, from 19 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open Portuguese LLM - OAB Exams | 33.53 | Accuracy (%) | 30.4 |
| Open Portuguese LLM - ENEM | 40.52 | Accuracy (%) | 30 |
| Open Portuguese LLM - ASSIN2 RTE | 78.58 | Macro F1 (%) | 29.6 |
| Open Portuguese LLM - BLUEX | 31.43 | Accuracy (%) | 27.2 |
| Open Chinese LLM - WinoGrande | 61.25 | Accuracy (%) | 25 |
| Open LLM Leaderboard - MMLU-Pro | 18.17 | Score | 19.8 |
| Open LLM Leaderboard - BBH | 34.57 | Score | 18.3 |
| Open Portuguese LLM - FaQuAD NLI | 43.97 | Macro F1 (%) | 17.1 |
| Open Chinese LLM - HellaSwag | 53.64 | Accuracy (%) | 16.3 |
| Open LLM Leaderboard - MuSR | 35.28 | Score | 14.2 |
| Open LLM Leaderboard - MATH Level 5 | 1.74 | Score | 12.5 |
| Open Chinese LLM - ARC Challenge | 43.17 | Accuracy (%) | 11.2 |
Interactive version: theaggregate.ai/model?slug=alma-13b-r · How It Works · Data refreshed daily, snapshot 2026-10-20.