GLM-5.5 后训练进展顺利,发布或已临近
智谱的 GLM-5.5 据说已经在后训练了,GLM-5.3 才出了两个月出头,感兴趣的话可以先关注一波动态。
推文称 GLM-5.5 已进入后训练阶段,而上一代 GLM-5.3 发布已超过两个月,推测新模型可能很快发布。文中还提到智谱等中国实验室当前普遍在投入 RSI(递归自我改进)方向的研究。
GLM-5.5 is already well advanced in the post-training phase, while GLM-5.3 is over two months old. Therefore, a release is expected soon.
What's interesting here, however, is how much all the Chinese labs are focusing on RSI. With all this concentrated effort, it's only a matter of time before RSI is truly achieved.