Greedy decoding

Also written: do_sample=False

Always choosing the most likely next token, so the same question gives the same answer every time.