The language model generating the candidate answers generates tokens until a full word is produced. The language models picking their answer choose the completion that results in the lowest perplexity independent of the tokenization.
1. perhaps even out of variants generated by other LLMs