Red Teaming Language Models to Reduce Harms: Methods, Scaling Behaviors, and Lessons Learned. (arXiv:2209.07858v2 [cs.CL] UPDATED)

We describe our early efforts to red team language models in order to
simultaneously discover, measure, and attempt to reduce their potentially
harmful outputs. We make three main contributions. First, we investigate
scaling behaviors for red teaming across 3 model sizes (2.7B, 13B, and 52B
parameters) and 4 model types: a plain language model (LM); an LM prompted to
be helpful, honest, and harmless; an LM with rejection sampling; and a model
trained to be helpful and harmless using reinforcement …

arxiv language language models lessons learned reduce scaling