Questions regarding algorithms performance in MineDojo and MineRl environment.

[Winston-Gu]

Dear authors,

Thank you so much for this amazing project! I believe this is one of the best codebases I have used, in terms of system design, consistent algorithm implementation, and detailed documentation.

I’m currently working on a project related to Minecraft agents, aiming to develop a model-based approach that enables agents to collect diamonds or perform a range of tasks. However, I’m still trying to understand the performance of different algorithms on Minecraft-related tasks.

With this in mind, I was wondering if you have any benchmarking results showing the performance of different algorithms on Minecraft tasks (MineRL or MineDojo)? I’d really appreciate it. Thanks again!

[belerico]

Hi @Winston-Gu, thank you for your kind words.
The Dreamer-V3 authors have extensively benchmarked various algorithms:

• https://www.nature.com/articles/s41586-025-08744-2
• https://arxiv.org/abs/2301.04104

Bear in mind that our current version of Dreamer-V3 does not correspond to the version published in their Nature paper.
We currently have an open issue (#271) to match the last version of Dreamer-V3.