arxiv preprint - Same Task, More Tokens: the Impact of Input Length on the Reasoning Performance of Large Language Models | AI Breakdown | Podwise