The provided oracle model and class for the Guess My City task only outputs yes or no, similar to the 20qs task. However, the provided dataset uses more natural language responses. Is this intentional? If so, are the experiments from the paper not using the natural language responses similar to those in the dataset?
The provided oracle model and class for the Guess My City task only outputs yes or no, similar to the 20qs task. However, the provided dataset uses more natural language responses. Is this intentional? If so, are the experiments from the paper not using the natural language responses similar to those in the dataset?