Cheap Reward Models for Small-Scale RLHF Experiments
Preference-based fine-tuning is expensive to study because reward models are large. We show that a reward model distilled to 1% of the size of the original recovers 94% of its ranking accuracy on held-out preference pairs, and that policies trained against the small model transfer to the large one with a modest loss. The result lowers the cost of reward-modelling experiments for groups with limited compute.