Llama-3-Instruct-8B-CPO-SimPO: benchmark results
Haoran Xu's CPO-SimPO fine-tune of Llama 3 8B Instruct, combining contrastive preference optimization with SimPO's length-normalized objective. Provider: haoranxu. Released 2024-06-19. Access: Open.
Aggregate Rating 1457 ± 17, rank #955 of 1656 rated models, from 24 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open Korean LLM - Ko-IFEval | 45.55 | Strict accuracy, prompt/instruction mean (%) | 89.8 |
| Open Korean LLM Leaderboard | 44.11 | Average Score (%) | 89.8 |
| Open Ko-LLM Leaderboard | 44.11 | Average (%) | 89.7 |
| Open Korean LLM - Ko-EQ-Bench | 37.11 | EQ-Bench score | 87.9 |
| Open Korean LLM - Ko-GSM8K (flexible extract) | 53.68 | Exact match, flexible extract (%) | 87.3 |
| Open LLM Leaderboard - IFEval | 70.46 | Score | 83.6 |
| Open Korean LLM - Ko-GPQA-Diamond | 25.25 | Normalized accuracy (%) | 75.9 |
| Open Korean LLM - KorNAT-CKA | 27.23 | Normalized accuracy (%) | 68.3 |
| Open Chinese LLM - TruthfulQA MC | 55.28 | Accuracy (%) | 62.1 |
| Open Chinese LLM - GSM8K | 47.92 | Accuracy (%) | 60.5 |
| Open LLM Leaderboard - MMLU-Pro | 36.86 | Score | 59.3 |
| Open Korean LLM - Ko-Winogrande | 61.01 | Accuracy (%) | 59.2 |
Interactive version: theaggregate.ai/model?slug=llama-3-instruct-8b-cpo-simpo · How It Works · Data refreshed daily, snapshot 2026-10-20.