Skip to content
Discussion options

You must be logged in to vote

是的,只能靠一些工程手段尽可能收敛,不过目前来看,模型的变量比工程影响更大,有些模型随机性比较大。另外分组环节用的是LLM,其实理想情况应该用 embedding 模型的,但是我们作为一个无服务的开源框架,无法做到依赖 embedding 模型服务。

Replies: 1 comment 1 reply

Comment options

You must be logged in to vote
1 reply
@lihaojin02
Comment options

Answer selected by lizhengfeng101
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Category
Q&A
Labels
None yet
2 participants