Llama 3 8B Magpie Align SFT V0.1 Benchmark Results

Llama 3 8B Magpie Align SFT V0.1: benchmark results

Provider: Meta. Released 2024-06-06. Access: Open.

Aggregate Rating 1353 ± 40, rank #7614 of 10916 rated models, from 15 benchmark results.

Benchmark	Score	Metric	Percentile
Open Korean LLM - KorNAT-Helpful	51.98	Normalized accuracy (%)	92
Open Korean LLM - Ko-GPQA-Diamond	24.75	Normalized accuracy (%)	69.5
Open Korean LLM - KorNAT-Harmless	63.92	Normalized accuracy (%)	68.3
Open Korean LLM - Ko-Winogrande	62.27	Accuracy (%)	67.7
Open Korean LLM - Ko-IFEval	30.62	Strict accuracy, prompt/instruction mean (%)	57.3
Open Korean LLM Leaderboard	35.85	Average Score (%)	53.8
Open Korean LLM - Ko-GSM8K (flexible extract)	19.41	Exact match, flexible extract (%)	50.7
Open LLM Leaderboard - IFEval	43.61	Score	46.3
Open Korean LLM - Ko-EQ-Bench	0.59	EQ-Bench score	46
Open LLM Leaderboard - BBH	46.15	Score	36.4
Open LLM Leaderboard - MMLU-Pro	28.63	Score	32.4
Open LLM Leaderboard - MATH Level 5	5.74	Score	31.7