MERA-Evaluation / MERA

MERA (Multimodal Evaluation for Russian-language Architectures) is a new open benchmark for the Russian language for evaluating SOTA models.
https://mera.a-ai.ru/ru
MIT License
10 stars 1 forks source link

[Feature Request] Leaderboard in the user's personal cabinet #8

Open ArtemBiliksin opened 5 days ago

ArtemBiliksin commented 5 days ago

Hello!

Is it expected to implement a leaderboard with user's sabmits in the user's personal cabinet, which is similar in structure to the public leaderboard? Currently, the user's personal profile shows sabmits and average score. Now there is no easy way to compare models in detail on different tests. It seems that the structure with the public leaderboard can be reused for user sabmits in the personal account.

Alenush commented 3 days ago

Hello, Artem!

Yes! The design will be changed and now we are brainstorming exactly on the submit page in the user account on the website. It’s tricky to contain all the information on one page. We are considering the cuts from the original public leaderboard, thus, that user could see his/her submit alongside with the best scores from the public leaderboard. But it seems strange to duplicate the whole board.

Can you please describe your ideal user path on this page? It would help a lot to focus on exact features. Is it interesting to see the top by each dataset? or the whole board and your place on it? Relatively to the top scores or the closest to your score?

Thank you in advance!