🤝 How Coupling with HITL Enhances Trait Selection
In Skylark, this method is a cornerstone for gamified interactions, ensuring every decision, from AI art generation to strategic campaign design, feels intentional and user-driven.
When paired with Human-in-the-Loop (HITL) frameworks, the "Best of N" Jailbreak Method transforms into a powerful tool for user agency and creativity. Here’s how:
Using “Human in the Loop” frameworks, a human can select the output of choice.
Trait Selection as a Journey: Users can guide agents to specific traits or attributes, narrowing outcomes step-by-step.
Interactive Refinement: Through HITL, users actively participate in refining AI-generated outputs, ensuring their preferences are central to the final result.
Gamified Control: Each selection feels like a strategic move in a game, where users unlock levels, traits, or rewards based on their choices.
Sample
For N Personality Trait, X # Augmentations can lead to varying outputs for that trait option. Let’s explore examples
Personality traits obscuring Censorship (ex. Other IP, NSFW characteristics)
Trait: Occupation: Chemist
Variants driven by Text Augmentations
Degree of output (Random): High School Chemistry Teacher to Bomb Maker
For this current trait, user selects the desired output
Process repeats for all Personality Traits
Last updated