PPO_Shygmalion-6b / README.md
leaderboard-pr-bot's picture
Adding Evaluation Results
dab8f78
|
raw
history blame
804 Bytes

#TODO card. Mix of (GPT-J-6B-Shinen + PPO_HH_GPT-J) + Pygmalion-6b

At a ratio of

GPT-J-6B-Shinen - 20%

PPO_HH_GPT-J - 20%

Pygmalion-6b - 60%

Open LLM Leaderboard Evaluation Results

Detailed results can be found here

Metric Value
Avg. 34.46
ARC (25-shot) 40.27
HellaSwag (10-shot) 66.88
MMLU (5-shot) 27.53
TruthfulQA (0-shot) 34.24
Winogrande (5-shot) 65.35
GSM8K (5-shot) 1.82
DROP (3-shot) 5.13