metadata
dataset_info:
features:
- name: model_id
dtype: large_string
- name: dataset_id
dtype: large_string
- name: problem_id
dtype: large_string
- name: original_problem
dtype: large_string
- name: permutation_type
dtype: large_string
- name: permutation_source
dtype: large_string
- name: base_accuracy
dtype: float64
- name: permuted_accuracy
dtype: float64
- name: absolute_accuracy_decay
dtype: float64
- name: relative_accuracy_decay
dtype: float64
- name: n_base_predictions
dtype: int64
- name: n_permuted_predictions
dtype: int64
- name: n_detrimental_permutations
dtype: int64
- name: permutations_causing_decay
dtype: large_string
- name: model_is_robust
dtype: bool
splits:
- name: validation
num_bytes: 233231
num_examples: 156
download_size: 51227
dataset_size: 233231
configs:
- config_name: default
data_files:
- split: validation
path: data/validation-*