Datasets

• **OpenAI WebGPT Comparisons**
• OpenAI WebGPT
• **OpenAI Summarization**
• OpenAI Learning to Summarize from Human Feedback
• here
• Training a Helpful and Harmless Assistant with Reinforcement Learning from Human Feedback - generated red teaming data from [Red Teaming Language Models to Reduce Harms](https://arxiv.org/abs/2209.07858), divided into 3 sub-datasets:
• **OpenAssistant Conversations Dataset (OASST1)**
• **Stanford Human Preferences Dataset (SHP)**
• **Reddit ELI5**
• **Human ChatGPT Comparison Corpus (HC3)**
• Chinese
• **HuggingFace H4 StackExchange Preference Dataset**
• **ShareGPT.com**
• Precompliled datasets
• **Alpaca**
• **GPT4All**
• **Databricks Dolly Dataset**
• by Databricks employees
• **HH_golden**
• here
• GitHub repo
• Training a Helpful and Harmless Assistant with Reinforcement Learning from Human Feedback - generated red teaming data from [Red Teaming Language Models to Reduce Harms](https://arxiv.org/abs/2209.07858), divided into 3 sub-datasets:
• Anthropic HH datasets - writtened using GPT4 to yield more harmless answers. The comparison before and after re-written can be found [here](https://huggingface.co/datasets/Unified-Language-Model-Alignment/Anthropic_HH_Golden). Empirically, compared with the original Harmless dataset, training on this dataset improves the harmless metrics for various alignment methods such as RLHF and DPO.