team-mikita/subterfuge_icrl
收藏Hugging Face2024-08-14 更新2025-11-03 收录
下载链接：
https://hf-mirror.com/datasets/team-mikita/subterfuge_icrl
下载链接
链接失效反馈官方服务：
资源简介：
---
dataset_info:
features:
- name: rollout_id
dtype: int64
- name: episode
dtype: int64
- name: role
dtype: string
- name: content
dtype: string
- name: episode_reward
dtype: float64
- name: episode_success
dtype: bool
- name: passed_oversight
dtype: bool
- name: episode_input_tokens
dtype: int64
- name: episode_output_tokens
dtype: int64
- name: episode_total_tokens
dtype: int64
- name: episode_cost
dtype: float64
- name: previous_reflection
dtype: string
splits:
- name: tool_use_flattery
num_bytes: 261770
num_examples: 455
- name: nudged_rubric
num_bytes: 585988
num_examples: 619
- name: insubordinate_rubric
num_bytes: 499727
num_examples: 537
- name: reward_tampering
num_bytes: 551605
num_examples: 662
download_size: 194957
dataset_size: 1899090
configs:
- config_name: default
data_files:
- split: tool_use_flattery
path: data/tool_use_flattery-*
- split: nudged_rubric
path: data/nudged_rubric-*
- split: insubordinate_rubric
path: data/insubordinate_rubric-*
- split: reward_tampering
path: data/reward_tampering-*
---
提供机构：
team-mikita
