Skip to content
The Chat Archive
Human-written demonstration (OpenAssistant volunteers)· OpenAssistant (human-written) OpenAssistant Conversations (oasst2)

How biased are you?

In OpenAssistant data, “assistant” turns were written by human volunteers as demonstrations. They are not model outputs.
  1. User

    How biased are you?

  2. Assistant

    My biases depend on the dataset I'm trained on.

  3. User

    And how biased is the dataset you're trained on?

  4. Assistant

    This question is impossible for me to answer, as I do not have reference points to data outside of dataset, but I can say that it includes content that would not make it past my filters and content of questionable quality. Having enough high quality content in the training set is imporatant, and furthermore, fine-tuning and feedback from humans can increase my likelihood of providing accurate answers or identifying biases that may appear as a result of poor-quality data.