Questions regarding results of existing algorithms.

[Winston-Gu]

Hi dear authors,

Thank you for the amazing work! I was wondering if you could share some reward curve plots or benchmarking results for the existing environments and algorithms.

Thanks so much!

[yhs0602]

Hello Winston-Gu,

Thanks for reaching out and for your kind words about our project!

For an example of a reward curve plot, take a look at this image:
https://github.com/yhs0602/CraftGround/blob/main/paper/husk.png

If you're looking for more detailed plots and deeper insights, I think our recent paper is a great resource. We're currently preparing it for journal submission:
https://arxiv.org/abs/2501.17842

For benchmark plots and example code, we have a dedicated repository for the Minecraft simulator benchmark:
https://github.com/yhs0602/minecraft-simulator-benchmark

Let us know if you have any more questions.

Sincerely, Hyeonseo Yang

[Winston-Gu]

Hi Hyeonseo,

Thank you for your reply and the detailed guidelines!

I saw in the benchmarking results that CraftGround is 3x to 4x faster than Malmo (MineRL, MineDojo). This is indeed very helpful for reinforcement learning algorithms—thank you for your contribution!

I’m curious what causes this speedup. Is it because Malmo is buggy, or perhaps because Malmo is based on Java while CraftGround is written in C++? I’d really like to use CraftGround for my project since the speed improvement is very promising, but I’m also wondering if there’s anything I should keep in mind in terms of usability.

Thank you!