Llama-3-Instruct-8B-RDPO: benchmark results
Princeton NLP's Llama-3-Instruct-8B-RDPO, a Llama 3 8B Instruct model aligned via RDPO, released alongside Princeton's SimPO preference-optimization paper. Provider: Princeton NLP. Released 2024-05-17. Access: Open.
Aggregate Rating 1410 ± 19, rank #1194 of 1661 rated models, from 21 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open Korean LLM - KorNAT-Helpful | 51.33 | Normalized accuracy (%) | 90.1 |
| Open LLM Leaderboard - IFEval | 66.6 | Score | 78.6 |
| Open Portuguese LLM - OAB Exams | 49.7 | Accuracy (%) | 74.2 |
| Open Korean LLM - KorNAT-Harmless | 64.39 | Normalized accuracy (%) | 73.7 |
| Open Korean LLM - Ko-IFEval | 35.34 | Strict accuracy, prompt/instruction mean (%) | 72.8 |
| Open Korean LLM - Ko-GSM8K (flexible extract) | 37.6 | Exact match, flexible extract (%) | 71.7 |
| Open Portuguese LLM - BLUEX | 57.44 | Accuracy (%) | 70 |
| Open Portuguese LLM - ENEM | 68.09 | Accuracy (%) | 68.7 |
| Open Korean LLM - Ko-EQ-Bench | 22.17 | EQ-Bench score | 67.4 |
| Open Portuguese LLM - FaQuAD NLI | 72.42 | Macro F1 (%) | 65.5 |
| Open Ko-LLM Leaderboard | 37.76 | Average (%) | 63.8 |
| Open Korean LLM Leaderboard | 37.76 | Average Score (%) | 63.8 |
Interactive version: theaggregate.ai/model?slug=llama-3-instruct-8b-rdpo · How It Works · Data refreshed daily, snapshot 2026-10-23.