7/29/2026 at 10:37:35 AM
Watching the video, 11:37 [1] made me feel very uneasy about the validity of those results. CMA-ES optimization times in seconds? That only works if you do no physics, no collision checks, no trajectory roll-outs. (Otherwise you're looking at 300+ iterations with 2000+ lanes each with 1000+ physics steps each, so millions to billions of simulation steps.)And that really makes me wonder, if 30 tokens for a robot can be so expressive that they replace 600 mio simulation steps?
Or the alternative would be that they just assume that the given trajectory is equally optimal for each robot, which to me seems like a HUUUUGE assumption. But if that's the case, then the results of this technique would be highly misleading, as they would tell you the robot is excellent for 1 example movement, while you wouldn't know that a tiny change to your example movement might make the situation much better overall with a different robot.
by fxtentacle
7/29/2026 at 1:24:58 PM
I'm also interested in how they figured out how to get the CMA-ES baseline to be this fast (less than 35 samples in Figure 5).Maybe they are warm-starting CMA-ES with a pre-trained controller (although at a glance it doesn't seem they do). Haven't read the paper yet though.
by dadoomer