Llama-3-Instruct-8B-SimPO-ExPO: benchmark results
chujiezheng's ExPO weight extrapolation of Princeton NLP's SimPO-tuned Llama 3 8B Instruct, nudging alignment beyond the original checkpoint. Provider: Meta. Released 2024-05-26. Access: Open.
Aggregate Rating 1470 ± 27, rank #2849 of 10916 rated models, from 13 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| AlpacaEval 2.0 | 45.78 | LC Win Rate (%) | 84.7 |
| Open LLM Leaderboard - IFEval | 64.34 | Score | 76.8 |
| Open Portuguese LLM - OAB Exams | 47.93 | Accuracy (%) | 67.9 |
| Open Portuguese LLM - BLUEX | 56.61 | Accuracy (%) | 66.6 |
| Open Portuguese LLM - ENEM | 64.87 | Accuracy (%) | 60.1 |
| Open Portuguese LLM - ASSIN2 RTE | 89.89 | Macro F1 (%) | 52.9 |
| Open LLM Leaderboard - MMLU-Pro | 34.01 | Score | 48.8 |
| Open Portuguese LLM - FaQuAD NLI | 58.73 | Macro F1 (%) | 42.7 |
| Open LLM Leaderboard - GPQA | 28.69 | Score | 41.8 |
| Open LLM Leaderboard - BBH | 47.65 | Score | 40.6 |
| WildBench | 35.02 | WB Score Task-Macro | 40.3 |
| Open LLM Leaderboard - MATH Level 5 | 7.02 | Score | 37.4 |
Interactive version: theaggregate.ai/model?slug=llama-3-instruct-8b-simpo-expo · How It Works · Data refreshed daily, snapshot 2026-11-08.