Llama-3-instruction-constructionsafety-layertuning: benchmark results
Provider: Meta. Access: Open.
Aggregate Rating 1361 ± 28, rank #7413 of 10916 rated models, from 15 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open Korean LLM - KorNAT-CKA | 37.4 | Normalized accuracy (%) | 97.1 |
| Open Korean LLM - Ko-Winogrande | 66.54 | Accuracy (%) | 88.5 |
| Open Korean LLM Leaderboard | 38.94 | Average Score (%) | 68.6 |
| Open Korean LLM - Ko-GPQA-Diamond | 24.24 | Normalized accuracy (%) | 62.4 |
| Open Korean LLM - Ko-GSM8K (flexible extract) | 27.14 | Exact match, flexible extract (%) | 61.1 |
| Open Korean LLM - Ko-EQ-Bench | 12.16 | EQ-Bench score | 58.1 |
| Open Korean LLM - Ko-IFEval | 30.8 | Strict accuracy, prompt/instruction mean (%) | 58.1 |
| Open LLM Leaderboard v1 - GSM8K | 33.59 | Accuracy (%) (5-shot) | 49.6 |
| Open LLM Leaderboard v1 - MMLU | 55.67 | Accuracy (%) (5-shot) | 36.9 |
| Open LLM Leaderboard v1 - ARC Challenge | 55.46 | Normalized accuracy (%) (25-shot) | 33.5 |
| Open LLM Leaderboard v1 - WinoGrande | 73.72 | Accuracy (%) (5-shot) | 33.3 |
| Open LLM Leaderboard v1 - TruthfulQA MC2 | 44.45 | MC2 (%) (0-shot) | 27.2 |
Interactive version: theaggregate.ai/model?slug=llama-3-instruction-constructionsafety-layertuning · How It Works · Data refreshed daily, snapshot 2026-11-08.