metadata
dataset_info:
features:
- name: input_text
dtype: string
- name: target_text
dtype: string
splits:
- name: train
num_bytes: 1680000
num_examples: 10000
- name: valid
num_bytes: 38912
num_examples: 1024
download_size: 662640
dataset_size: 1718912
Dataset Card for "circle.10.9.10.10000"
Data examples for the paper Roll the dice & look before you leap: Going beyond the creative limits of next-token prediction