[staticman2]

Maybe it depends on the model but I find you'll get a different result if you say "write a dialog in which, A, B, and C talk about D" versus "read what A said and reply as B". The latter will result in each participant talking longer.

[webappguy]

Not sure talking longer is the goal. More so, the focus and separation of each facilitates a interplay and dynamic with which an attention window on a segmented linear response (be A, B and C) cannot be individually represent nearly as rhobustly (the main inference is the primary focus). Would love to hear some other opinions chime in here.

[webappguy]

additionally, each agent can be a model using its own RAG or training data as well