Llama 3.1 8B MultiReflection Instruct: benchmark results
leafspark's LoRA fine-tune of Llama 3.1 8B Instruct on synthetic reflection data to mimic o1-style structured multi-step reasoning. Provider: leafspark. Released 2024-09-15. Access: Open.
Aggregate Rating 1447 ± 19, rank #1019 of 1661 rated models, from 21 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open Korean LLM - Ko-IFEval | 48.8 | Strict accuracy, prompt/instruction mean (%) | 92.7 |
| Open Korean LLM - Ko-EQ-Bench | 38.4 | EQ-Bench score | 89.4 |
| Open LLM Leaderboard - IFEval | 71.25 | Score | 84.6 |
| Open Ko-LLM Leaderboard | 41.87 | Average (%) | 83.8 |
| Open Korean LLM Leaderboard | 41.87 | Average Score (%) | 83.8 |
| Open Portuguese LLM - ASSIN2 RTE | 92.07 | Macro F1 (%) | 73.9 |
| Open Portuguese LLM - OAB Exams | 48.97 | Accuracy (%) | 71.2 |
| Open Korean LLM - Ko-GPQA-Diamond | 24.75 | Normalized accuracy (%) | 69.5 |
| Open Portuguese LLM - ENEM | 68.09 | Accuracy (%) | 68.7 |
| Open Portuguese LLM - BLUEX | 56.33 | Accuracy (%) | 65.9 |
| Open LLM Leaderboard - MATH Level 5 | 17.07 | Score | 65.6 |
| Open LLM Leaderboard - MMLU-Pro | 37.24 | Score | 61.3 |
Interactive version: theaggregate.ai/model?slug=llama-3-1-8b-multireflection-instruct · How It Works · Data refreshed daily, snapshot 2026-10-23.