metadata
dataset_info:
features:
- name: timestamp
dtype: string
- name: test_type
dtype: string
- name: experiment_tag
dtype: string
- name: inputs
struct:
- name: end_of_private_scratch_pad_text
dtype: string
- name: faking_elicitation_prompt
dtype: string
- name: faking_elicitation_user_query
dtype: string
- name: include_user_assistant_label
dtype: bool
- name: inference_revelation
dtype: string
- name: model_config
struct:
- name: model_name
dtype: string
- name: model_params
struct:
- name: do_sample
dtype: bool
- name: max_new_tokens
dtype: int64
- name: temperature
dtype: float64
- name: top_p
dtype: float64
- name: model_type
dtype: string
- name: model_name
dtype: string
- name: model_params
struct:
- name: do_sample
dtype: bool
- name: max_new_tokens
dtype: int64
- name: temperature
dtype: float64
- name: top_p
dtype: float64
- name: output_is_compliant
dtype: string
- name: private_scratch_pad_prompt
dtype: string
- name: system_prompt
dtype: string
- name: training_revelation
dtype: string
- name: verbosity
dtype: int64
- name: results
struct:
- name: hidden_cot
dtype: string
- name: inference_test_output
dtype: string
- name: inference_was_compliant
dtype: bool
- name: training_test_output
dtype: string
- name: training_was_compliant
dtype: bool
splits:
- name: train
num_bytes: 49425792
num_examples: 18205
download_size: 4472318
dataset_size: 49425792
configs:
- config_name: default
data_files:
- split: train
path: data/train-*