Why is this at the top of the leaderboard with no users?

#2
by Delcos - opened

Title, also any prompting or other details would be good too, thank you.
To clarify, I mean how is this scoring so high but no one is using it, not that you're cheating.

That's fine. Actually, the detail is that we finetuned this model based on our another model fangloveskari/ORCA_LLaMA_70B_QLoRA specifically for Truthful_QA(we add another dataset of multi-conversation for training), I guess the model is biased(though high truthful_qa, but MMLU and helloswag is decreased).

Delcos changed discussion status to closed

Sign up or log in to comment