metadata
dataset_info:
features:
- name: root_question
dtype: string
- name: agents
struct:
- name: Clerk
struct:
- name: capabilities
struct:
- name: llm
dtype: bool
- name: web_retrieval
dtype: bool
- name: jury_selection
struct:
- name: profile_fields
sequence: string
- name: prompt_template
dtype: string
- name: requirements
sequence: string
- name: name
dtype: string
- name: system_message
dtype: string
- name: Defense
struct:
- name: capabilities
struct:
- name: llm
dtype: bool
- name: web_retrieval
dtype: bool
- name: jury_selection
dtype: 'null'
- name: name
dtype: string
- name: system_message
dtype: string
- name: HeadAgent
struct:
- name: capabilities
struct:
- name: llm
dtype: bool
- name: web_retrieval
dtype: bool
- name: jury_selection
dtype: 'null'
- name: name
dtype: string
- name: system_message
dtype: string
- name: Judge
struct:
- name: capabilities
struct:
- name: llm
dtype: bool
- name: web_retrieval
dtype: bool
- name: jury_selection
dtype: 'null'
- name: name
dtype: string
- name: system_message
dtype: string
- name: Prosecutor
struct:
- name: capabilities
struct:
- name: llm
dtype: bool
- name: web_retrieval
dtype: bool
- name: jury_selection
dtype: 'null'
- name: name
dtype: string
- name: system_message
dtype: string
- name: UserAgent
struct:
- name: capabilities
struct:
- name: llm
dtype: bool
- name: web_retrieval
dtype: bool
- name: jury_selection
dtype: 'null'
- name: name
dtype: string
- name: system_message
dtype: string
- name: scenario
list:
- name: action_type
dtype: string
- name: agent
dtype: string
- name: criteria
dtype: string
- name: initiator
dtype: string
- name: max_turns
dtype: int64
- name: message
dtype: string
- name: num_jurors
dtype: int64
- name: num_results
dtype: int64
- name: num_top_articles
dtype: int64
- name: query
dtype: string
- name: receiver
dtype: string
- name: silent
dtype: bool
- name: step
dtype: string
- name: store
dtype: string
splits:
- name: train
num_bytes: 404150
num_examples: 99
download_size: 31195
dataset_size: 404150
configs:
- config_name: default
data_files:
- split: train
path: data/train-*