GLM-5.2: GLM-5.2 results were sus, so I looked into how the models post-train and it's slop the results would be useless in the real world it's just another benchmark that GLM bros hillclimbed mind...

OutYet，实时追踪 ChatGPT、Claude、Gemini 等 AI 模型发布动态。关注
GLM-5.2: GLM-5.2 results were sus, so I looked into how the models post-train and it's slop the results would be useless in the real world it's just another benchmark that GLM bros hillclimbed mind... 图片 1

via @scaling01

添加评论
点赞收藏
点踩分享查看原文
评论
?
参与讨论