Quick multi-agent with Nemo investigation results (creative writing)
#2
by
Lambent
- opened
For anyone using this as part of a multi-agent setup with a Mistral Nemo model, I tried having both this model and arsenic-nemo-unleashed-12B generate a scene and then write a scene using those two drafts as the input. (Q4_K_M)
Results seem to be that arsenic-nemo is more skilled at smoothly integrating output from this model into its response than the other way around. Eidolon changes their final response much less from its initial response, where arsenic-nemo writes fairly distinctly from both.