Spaces:

---

AxiomicLabs

/

Open_SLM_Leaderboard

like 78

Running

Hugging Face Model Page: https://huggingface.co/veyra-ai/Veyra2-Apricot-50M-Base

Zero-shot results:

PIQA acc_norm: 62.02%
ARC-Easy acc_norm: 42.47%
ARC-Challenge acc_norm: 23.38%
HellaSwag acc_norm: 31.25%
Winogrande acc: 50.12%
OpenBookQA acc_norm: 28.80%
BoolQ acc: 60.67%
ArithMark-2.0 acc: 29.00%

Also, sorry for missing your question about adding Veyra2 30M in a different discussion. I did not see it at the time. I do not think Veyra2 30M is worth adding now because its results were poor, and Veyra 30M effectively replaces it after I converted Veyra 30M to a standard Llama config.