为什么大型语言模型没有用它们自己的“思维链”进行训练？
1 分•作者: simianwords•4 个月前
我意识到，如果不允许推理代币，模型表现会非常差。它无法进行简单的算术或逻辑运算，并且会产生一些幻觉。<p>但通过允许它稍微思考一下再回答，结果会好得多，也更值得信赖。<p>这展示了一个干净的强化学习环境……或者只是一个不错的数据集。在这个环境中，你对模型进行两次提示——一次不允许思考，一次允许思考。如果“不思考”的结果与“思考”后得到的结果相矛盾，则对“不思考”的结果进行惩罚。
查看原文
I realise that without allowing reasoning tokens, a model performs very poorly. It can't perform simple arithmetic or simple logic and hallucinates a bit.<p>But by allowing it to think a bit and then answer, the result is much better and way more trustable.<p>This shows a clean RL environment.. or just a nice data-set. Where you prompt the model two times - one without allowing thinking and one with thinking. Penalise the result from non thinking if the result contradicts the answer obtained from thinking.