K-Scale Humanoid Benchmark Leaderboard

Welcome to the K-Scale Labs Humanoid Benchmark Leaderboard!

What is this?

This leaderboard is for people who want to try training their own RL policies and running them on a K-Scale robot.

How do I win?

We will evaluate based on how cool the robot looks when it runs your code, for some subjective measurement of "cool". All submissions will get a spot on the leaderboard, and we will be grateful for your assistance in helping us move towards our mission.

We will roughly score the submissions using an ELO rating system, since "coolness" should be roughly monotonic and is easiest to evaluate pair-wise. If you wish to contest your score or someone else's score, please tell us in our Discord.

How do I get started?

  1. Create a new repository from this template.
  2. Train your own policy and make sure that you can run it in kos-sim, following the instructions in the template repository.
  3. After you have a policy which runs in kos-sim using the deploy.py script, go to our Discord and submit a link with your policy to the "【🧠】submissions" channel
  4. One of us will run your policy on the real robot and add the results to the leaderboard. We will try to run new policies every night, so you shouldn't have to wait too long to see your policy on the leaderboard.

Discord · K-Scale Documentation · Leaderboard Github

Submissions

Name Submitter Date Submission Sim Video Real Video
MLP Baseline kscalelabs 2025-04-25 Submission Sim Video Real Video