Lumo censorship — still too restrictive for creative writing
Hi Proton team,
Thanks for the September update loosening Lumo's filters — it's a big improvement.
However, I'm still hitting refusals on adult creative writing between fictional characters (no minors, no non-consent, no illegal content — just explicit fiction/roleplay between consenting adults). The model occasionally breaks character or adds disclaimers mid-scene, which kills the experience.
Two things that would help:
Clarify in the system prompt that explicit fiction between adult fictional characters is acceptable, so the model doesn't second-guess itself mid-generation.
If possible, route creative/roleplay conversations to a model with lighter guardrails, or fine-tune the current ones to be more permissive on this specific use case.
I understand you need boundaries for safety, but "adult fiction" and "harmful content" aren't the same thing. The current line is drawn a bit too far.
Happy to elaborate if useful.