2026-04-01-llm #252
Replies: 4 comments 2 replies
-
|
图是用claude画的 |
Beta Was this translation helpful? Give feedback.
1 reply
-
|
感谢博主讲解,内容上基本把主流工作模式和每个节点的目标和难点都讲解了,很到位!。 如果从传统 LLM 训练 -> LLM + 对齐人类偏好 -> 以agent 的LLM, 讲解似乎更好。如果全文再贯穿某一个团队的持续工作(比如deepseek, openai,anthropic等) 讲解会更直观(个人观点,个人建议)~ |
Beta Was this translation helpful? Give feedback.
1 reply
-
|
配图都好好看!(有啥提示词可参考的嘛);整体内容量很扎实,收获很大! |
Beta Was this translation helpful? Give feedback.
0 replies
-
|
感谢博主讲解,而且流程图很好看!请问是用哪个skill画的吗? |
Beta Was this translation helpful? Give feedback.
0 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
-
2026-04-01-llm
很多人会把模型升级理解为参数变大,但线上体感差异常常出在后半段训练和发布链路。这篇从预训练一路讲到蒸馏上线,重点看数据工程、系统配方、后训练、评测奖励和 Agent 训练怎么一起影响最终表现。最后会看到,模型变强通常是权重、训练链路和部署决策共同作用的结果,不只是参数规模。
https://tw93.fun/2026-04-03/llm.html
Beta Was this translation helpful? Give feedback.
All reactions