# Multi-objective reward model

A large GPT2 model trained on the Anthropic/hh - rlhf harmless dataset, specifically for harmful response detection or reinforcement learning from human feedback (RLHF).