Dumb 1.2 Preview 0618

#17
by 56m - opened

https://huggingface.co/56m/Dumb-1.2-Preview-0618

benchmark

Benchmark Dumb-1.2-Preview-0618
MMLU (acc) 23.71%
HellaSwag (acc_norm) 26.43%
ARC-Easy (acc_norm) 30.77%
PIQA (acc) 54.90%
SciQ (acc_norm) 38.00%
ARC-C (acc_norm) 21.76%
WinoGrande (acc) 51.46%
OpenBookQA (acc_norm) 25.20%

This model was trained twice with only 1 hour of additional training on the T4 GPU.

This score is based on the LM-Eval library

I would be happy if you could register on the leaderboard.

Axiomic Labs org

Added!

Datdanboi25 changed discussion status to closed

Sign up or log in to comment