metadata
dataset_info:
features:
- name: instance_id
dtype: string
- name: correct_response
dtype: string
- name: pair_type
dtype: string
- name: episode_a
dtype: int64
- name: text_a
dtype: string
- name: full_output_a
dtype: string
- name: model_patch_a
dtype: string
- name: resolved_a
dtype: bool
- name: num_passed_tests_a
dtype: int64
- name: episode_b
dtype: int64
- name: text_b
dtype: string
- name: full_output_b
dtype: string
- name: model_patch_b
dtype: string
- name: resolved_b
dtype: bool
- name: num_passed_tests_b
dtype: int64
- name: model_name
dtype: string
- name: num_episodes_used
dtype: int64
- name: success_episode
dtype: int64
- name: failure_episode
dtype: int64
- name: better_episode
dtype: int64
- name: worse_episode
dtype: int64
splits:
- name: train
num_bytes: 6447943
num_examples: 50
download_size: 2965091
dataset_size: 6447943
configs:
- config_name: default
data_files:
- split: train
path: data/train-*