I guess one way you could do that is create more stop sequences like "for Example:" and then split off mini language models and to generate each segment. Althou

I guess one way you could do that is create more stop sequences like "for Example:" and then split off mini language models and to generate each segment. Although I'm not sure how you would not get duplicates.