I'm using XGBoost version 3.0.0 to train a Learning to Rank model with the rank:ndcg objective. I would like to train the model using the non-exponential (linear gain) version of the nDCG listwise loss, so I set the parameter:
ndcg_exp_gain = False

I’m posting to confirm a couple of things:

1. Is setting ndcg_exp_gain=False the only change required to switch from exponential to non-exponential (linear gain) nDCG loss during training?
2. Are there any known caveats, limitations, or gotchas associated with using ndcg_exp_gain=False that I should be aware of?

Any clarification or best practices would be greatly appreciated!