[2209.07858] Red Teaming Language Models to Reduce Harms: Methods, Scaling Behaviors, and Lessons Learned

Abstract: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 learning from human feedback (RLHF). We find that the RLHF models are increasingly difficult to red team as they scale, and we find a flat trend with scale for the other model types. Second, we release our dataset of 38,961 red team attacks for others to analyze and learn from. We provide our own analysis of the data and find a variety of harmful outputs, which range from offensive language to more subtly harmful non-violent unethical outputs. Third, we exhaustively describe our instructions, processes, statistical methodologies, and uncertainty about red teaming. We hope that this transparency accelerates our ability to work together as a community in order to develop shared norms, practices, and technical standards for how to red team language models.

Abstract: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 learning from human feedback (RLHF). We find that the RLHF models are increasingly diffic

related reading

• Red Teaming Language Models with Language Models — Google DeepMinddeepmind.com
• Discovering Language Model Behaviors with Model-Written Evaluations — LessWronglesswrong.com
• Transformer Circuits Threadtransformer-circuits.pub
• [2402.04249] HarmBench: A Standardized Evaluation Framework for Automated Red Teaming and Robust Refusalarxiv.org
• [2312.06942] AI Control: Improving Safety Despite Intentional Subversionarxiv.org
• Defining LLM Red Teaming | NVIDIA Technical Blogdeveloper.nvidia.com
• Externalized reasoning oversight: a research direction for language model alignment — AI Alignment Forumalignmentforum.org
• Mediumai-alignment.com
• Automated Alignment Researchers: Using large language models to scale scalable oversight \ Anthropicanthropic.com
• [2203.02155] Training language models to follow instructions with human feedbackarxiv.org
• Surfacing Pathological Behaviors in Language Models | Transluce AItransluce.org
• Lapis Labslapis.rocks