cypher rlhf & multiverse
I currently work on the Cypher project, where I write prompts for specific categories, evaluate two different model responses, and select the best one. I provide comments explaining which response is better and why, based on predefined criteria.