Create adversarial counterparts to each generative stage
Ideas
Rather than having it analyse the whole output, it should...
- Drill down to identify if there are agents below it
- If yes, trust their analyses and only run inter-unit coherence checks. If no, check content
It should probably be handed some instructions from parent as well about immediate siblings
Create adversarial counterparts to each generative stage
Ideas
Rather than having it analyse the whole output, it should...
It should probably be handed some instructions from parent as well about immediate siblings