Suspicious benchmark scores
#1
by ProCreations - opened
The model seems very benchmaxxed. Can we see scores on benchmarks like DeepSWE? The benchmarks results you show are way too good to be true.
Agreed, very suspicious, will most definitely need 3rd party testing
HLE got to 99.44%... Do you think I believe that? “Accidentally” mixed the test set into the training set, right? 😆
I think it's a joke definitely🤣
Please report this repository, this should not be on HF. Damaging
Please report this repository, this should not be on HF. Damaging
LOL
Check the video. I have no clue why people are so rude about it lol