NeuralLLaMa-3-8B-ORPO-v0.3: benchmark results
Kukedlc's ORPO fine-tune of Llama 3 8B Instruct, preference-trained on mlabonne's orpo-dpo-mix-40k dataset. Provider: Other. Released 2024-05-01. Access: Open.
Aggregate Rating 1414 ± 15, rank #1174 of 1656 rated models, from 52 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open LLM Leaderboard v1 - GSM8K | 72.93 | Accuracy (%) (5-shot) | 98.3 |
| Open CoT - LogiQA 2 | 16.48 | CoT Gain (%) | 96.9 |
| Open CoT - LogiQA | 10.22 | CoT Gain (%) | 95.4 |
| Open CoT Leaderboard | 14.52 | Average CoT Gain (%) | 94.7 |
| Open CoT - LSAT Logical Reasoning | 20.2 | CoT Gain (%) | 93.1 |
| Open LLM Leaderboard v1 - MMLU | 68.39 | Accuracy (%) (5-shot) | 91.3 |
| Open Medical LLM - PubMedQA | 77 | Accuracy (%) | 89.3 |
| Open Medical LLM - MedMCQA | 58.52 | Accuracy (%) | 89.1 |
| Open CoT - LSAT Reading Comprehension | 20.07 | CoT Gain (%) | 85.9 |
| Open Medical LLM - MMLU College Biology | 79.86 | Accuracy (%) | 84.4 |
| Open LLM Leaderboard v1 - ARC Challenge | 69.54 | Normalized accuracy (%) (25-shot) | 83.5 |
| Open Korean LLM - Ko-GPQA-Diamond | 25.76 | Normalized accuracy (%) | 81.3 |
Interactive version: theaggregate.ai/model?slug=neuralllama-3-8b-orpo-v0-3 · How It Works · Data refreshed daily, snapshot 2026-10-20.