G.A.R.Y. (Guided Artificially Resourceful Yes-man)

A clone of Dolly (https://github.com/databrickslabs/dolly)

Trained on 8xa100s over the course of 45 minutes. (total time less than 3 hours with false starts and getting less optimal results while learning how best to proceed.)

license: cc-by-nc-2.0

Open LLM Leaderboard Evaluation Results

Detailed results can be found here

Metric	Value
Avg.	35.1
ARC (25-shot)	41.3
HellaSwag (10-shot)	65.97
MMLU (5-shot)	26.78
TruthfulQA (0-shot)	37.91
Winogrande (5-shot)	64.72
GSM8K (5-shot)	0.91
DROP (3-shot)	8.1