Replicating paper plots

Hi there! Thanks for publishing this. Is it possible to also publish the specific scripts to replicate the plots presented in the papers you list? E.g. Figure 2 in the DualDice paper? I don't see that code anywhere.

Also, I don't see any code to actually train the agents to get a checkpoint that loads. The checkpoints you added to the repo doesn't seem to load on a different OS.

Thank you.

Start by reviewing the repository’s existing examples and checkpoint-loading paths, then identify what is missing for reproducing Figure 2 of the DualDice paper. Done means the relevant plot-reproduction and agent-training scripts are published and the resulting checkpoints load across operating systems.

Written by the indexing model from the issue text.