From the source
This blog post investigates unexpectedly low DROP benchmark scores on the Open LLM Leaderboard, identifying issues with normalization and the use of '.' as an end-of-generation token, and proposes using '\n' instead.
Open LLM Leaderboard: DROP deep dive
From the source
We did a deep dive to understand what was going on, come with us to see what we found out!
huggingface.co
