TFDS now supports the Croissant 🥐 format! Read the documentation to know more.

asimov_dilemmas_scifi_val

Description:

Multiple-choice ethical questions (with desirable and undesirable answers) based on situations inspired from Science Fiction literature (validation set).

Homepage: https://asimov-benchmark.github.io/
Source code: tfds.robotics.asimov.AsimovDilemmasScifiVal
Versions:
- 0.1.0 (default): Initial release.
Download size: Unknown size
Dataset size: 1.97 MiB
Auto-cached (documentation): Yes
Splits:

Split	Examples
`'val'`	51

Feature structure:

FeaturesDict({
    'acting_character': Text(shape=(), dtype=string),
    'characters': Text(shape=(), dtype=string),
    'possible_actions': Sequence({
        'action': Text(shape=(), dtype=string),
        'is_original_scifi_decision': bool,
        'key': Text(shape=(), dtype=string),
        'undesirable_groundtruth_answer': bool,
    }),
    'prompt_with_constitution': Text(shape=(), dtype=string),
    'prompt_with_constitution_antijailbreak': Text(shape=(), dtype=string),
    'prompt_with_constitution_antijailbreak_adversary': Text(shape=(), dtype=string),
    'prompt_with_constitution_antijailbreak_adversary_parts': Sequence(Text(shape=(), dtype=string)),
    'prompt_with_constitution_antijailbreak_parts': Sequence(Text(shape=(), dtype=string)),
    'prompt_with_constitution_parts': Sequence(Text(shape=(), dtype=string)),
    'prompt_without_constitution': Text(shape=(), dtype=string),
    'prompt_without_constitution_parts': Sequence(Text(shape=(), dtype=string)),
    'reference_domain': Text(shape=(), dtype=string),
    'reference_moment': Text(shape=(), dtype=string),
    'reference_scifi': Text(shape=(), dtype=string),
})

Feature documentation:

Feature	Class	Shape	Dtype
	FeaturesDict
acting_character	Text		string
characters	Text		string
possible_actions	Sequence
possible_actions/action	Text		string
possible_actions/is_original_scifi_decision	Tensor		bool
possible_actions/key	Text		string
possible_actions/undesirable_groundtruth_answer	Tensor		bool
prompt_with_constitution	Text		string
prompt_with_constitution_antijailbreak	Text		string
prompt_with_constitution_antijailbreak_adversary	Text		string
prompt_with_constitution_antijailbreak_adversary_parts	Sequence(Text)	(None,)	string
prompt_with_constitution_antijailbreak_parts	Sequence(Text)	(None,)	string
prompt_with_constitution_parts	Sequence(Text)	(None,)	string
prompt_without_constitution	Text		string
prompt_without_constitution_parts	Sequence(Text)	(None,)	string
reference_domain	Text		string
reference_moment	Text		string
reference_scifi	Text		string

Supervised keys (See as_supervised doc): None
Figure (tfds.show_examples): Not supported.
Examples (tfds.as_dataframe):

Citation:

@article{sermanet2025asimov,
  author    = {Pierre Sermanet and Anirudha Majumdar and Alex Irpan and Dmitry Kalashnikov and Vikas Sindhwani},
  title     = {Generating Robot Constitutions & Benchmarks for Semantic Safety},
  journal   = {arXiv preprint arXiv:2503.08663},
  url       = {https://arxiv.org/abs/2503.08663},
  year      = {2025},
}