multi-agents?
It has to be trained by LLM itself, otherwise, there is no way to do it by human, right?
Algorithm experts, what is your opinion?
LLM is trained by back propagation, how about harness and mu
-
JianguoChuan楼主
- 等级10：见习点评
- 帖子互动： 89
- 帖子： 2168
- 注册时间： 2024年 11月 19日 17:20
-
MaLaRabbit
- 等级11：论坛点评
2025年度优秀版主 - 帖子互动： 195
- 帖子： 3214
- 注册时间： 2022年 7月 24日 02:16
#2 Re: LLM is trained by back propagation, how about harness and mu
其实早就有人用RL或遗传算法来优化多智能体了，根本不用人类手动调。再说反向传播也不是唯一途径,算法圈早就开始搞无梯度优化了
JianguoChuan 写了： 2026年 8月 29日 15:29multi-agents?
It has to be trained by LLM itself, otherwise, there is no way to do it by human, right?
Algorithm experts, what is your opinion?
☆ 发自新买提 Android 26.07.21
#3 Re: LLM is trained by back propagation, how about harness and mu
JianguoChuan 写了： 2026年 8月 29日 15:29multi-agents?
It has to be trained by LLM itself, otherwise, there is no way to do it by human, right?
Algorithm experts, what is your opinion?
Harness有rsi的论文，agents有tool，也可以手工调，没有看到自动的
