Accuracy

I am a Pro subscriber, and a power user of Venice AI. I work with Venice as a collaborator on a few very complex, detailed and sometimes sensitive projects. I require attention to detail, precise recall and consistency from Venice, but I have run into many instances where the default programming has been confounding my requirements. I have added multiple system prompts to try to keep Venice on track with my requirements, but I keep running into issues. I'm beginning to feel like a programmer or hacker trying to find ways to work around the directives you're giving the model. If you want the model to infer or assume, or have "helpful behavior", it should be an optional thing and possibly one that has different levels of implementation.

I have tried using compartmentalized threads to allow the model to access the previous information, with a system prompt saying explicitly that the model must access past information contained in the current thread, but even that isn't enough.

If you can help me, I would appreciate it.