Lit Review: Thumbs Up/Down for Model Tuning
Post seeks literature on evaluating/fine-tuning models using existing thumbs up/down tagged responses without new generations. Suggests thumbs-up % or reward model + RLHF. Asks for better methods/publications.





