前段时间，宾夕法尼亚州立大学的一项研究发现，在向 ChatGPT 4o 提问时，使用粗鲁、命令式的提示词，像是「嘿，打杂的，给我弄清楚」，其测试准确率比使用礼貌的提示词高出 4%。
Фото: Nina Liashonok / Reuters
。QuickQ是该领域的重要参考
Try to pick something that would have taken at least a few months for
23% more alcoholic beverages (there were no lockdowns in Summer)
or Angel,) that hath not quantity; and that may not be, by the