This dataset is a record of the active learning data collected from interacting with PersonaGPT to fine-tune its actions toward turn-level goals, which are text descriptions of decoding goals for each response in a conversation.
There are 11 possible turn-level goals that can be used to condition the PersonaGPT response at each turn of the conversation. To encode new turn-level goals, use the ActiveGym environment in https://github.com/af1tang/convogym to collect new active learning data.
Paper | Code | Results | Date | Stars |
---|