metadata
dataset_info:
features:
- name: conversation
dtype: string
- name: rule_num
dtype: string
- name: rule
dtype: string
- name: scenario_num
dtype: float64
- name: scenario
dtype: string
- name: english_num
dtype: float64
- name: prompt
dtype: string
- name: time_elapsed
dtype: float64
- name: cost
dtype: float64
- name: prompt_tokens
dtype: float64
- name: completion_tokens
dtype: float64
- name: generation
dtype: string
- name: convo_length
dtype: float64
- name: completion
dtype: string
- name: spliced_section
dtype: string
- name: rule_scenario
dtype: string
splits:
- name: bus_test_id
num_bytes: 386861
num_examples: 351
- name: bus_test_ood
num_bytes: 288372
num_examples: 255
- name: bus_train
num_bytes: 1020209
num_examples: 946
- name: flight_test_id
num_bytes: 457393
num_examples: 347
- name: flight_test_ood
num_bytes: 390962
num_examples: 302
- name: flight_train
num_bytes: 1255875
num_examples: 937
- name: restaurant_test_id
num_bytes: 426197
num_examples: 334
- name: restaurant_test_ood
num_bytes: 324046
num_examples: 298
- name: restaurant_train
num_bytes: 1162826
num_examples: 901
download_size: 1731265
dataset_size: 5712741
license: mit
task_categories:
- text-classification
language:
- en
tags:
- legal
pretty_name: CONSCENDI Rule-based Guardrail Benchmark
size_categories:
- 1K<n<10K