Model leaderboard page comparing fine-tuned checkpoints

[SahilKumar75]

No UI exists to compare eval metrics (perplexity, ROUGE, accuracy) across multiple fine-tuned checkpoints side by side.

[pal404error]

For the leaderboard page, one option is to pull from an existing open ranking instead of maintaining your own from scratch. ModelRank publishes its full ranking as CSV and JSON and the weights are public, so it is easy to reuse. https://rankmodel.github.io