Beyond Semantics: The Unreasonable Effectiveness of Reasonless Intermediate Tokens 事件

Beyond Semantics: The Unreasonable Effectiveness of Reasonless Intermediate Tokens arXiv:2505.13775v4 Announce Type: replace-cross Abstract: Recent impressive results from large reasoning models have been interpreted as a triumph of Chain of Thought (CoT), and especially of the process of training on CoTs sampled from base LLMs in order to help find new reasoning patterns. While these traces certainly seem to help model performance, it is not clear how they influence it, with some works ascribi