Supporting the inclusion of specialized judge models from the Hub

We have 2k votes now and an interesting leaderboard is shaping up with proprietary models and leading open-source models. We'd like to include CompassJudger-1, Flow-judge, and Prometheus 2 models in the Judge Arena to see how specialized judges stack up. To support this effort, we would really appreciate GPU / storage resources 🙏
We currently support the inference of existing models on the leaderboard out of our own pocket