This checkpoint has been reproduced based on the code provided in the
facebookresearch/coconut repository and the experimental settings described in the paper
Training Large Language Models to Reason in a Continuous Latent Space. Please refer to these sources for further details on the methodology and configuration used in this experiment.
