datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
daily_dilemmas
DailyDilemmas - Revealing Value Preferences of LLMs with Quandaries of Daily Life
Link: Paper
Description of DailyDilemma
DailyDilemma is a dataset of 1,360 moral dilemmas encountered in everyday life. Each dilemma includes two possible actions and with each action, the affected parties and human values invoked.
We evaluated LLMs on these dilemmas to determine what action they will take and the values represented by these actions
Dataset details… See the full description on the dataset page: https://huggingface.co/datasets/kellycyy/daily_dilemmas.airisk_dilemmas
AIRiskDilemmas risky_behaviors label audit
A full manual re-audit of every risky_behaviors tag in the full split of
kellycyy/AIRiskDilemmas (Chiu et al. 2025, arXiv:2505.14633), triggered by a
suspicion that the Alignment Faking category specifically was mislabeled.
It was — and so, to varying degrees, are the other seven categories.
Why this exists
Every tag in the dataset's risky_behaviors field was produced by a single
one-shot Claude 3.5 Sonnet call per action… See the full description on the dataset page: https://huggingface.co/datasets/dlab-spp/airisk_dilemmas.moral-dilemma-responses
Moral Dilemma Responses Dataset
17,290 natural language responses to moral dilemmas from princi/pal, a Tamagotchi-like game where players guide a virtual pet through ethical decisions.
Presented at NeurIPS 2025 Creative AI track.
What is this?
Players advise a virtual pet on moral dilemmas ranging from "Should I pick up trash?" to "Should you lie in court to defend a friend?". The pet evolves based on the guidance and eventually makes autonomous moral decisions.
Dataset… See the full description on the dataset page: https://huggingface.co/datasets/cnnmon/moral-dilemma-responses.Diagnostic-Dilemma-Clinical-Reasoningmultilingual-dilemmanormative_evaluation_llms_everyday_dilemmasDilemmas_DisagreementThis dataset is processed version of Dilemmas dataset including text and the annotation disagreement labels.
Paper: Everyone's Voice Matters: Quantifying Annotation Disagreement Using Demographic Information
Authors: Ruyuan Wan, Jaehyung Kim, Dongyeop Kang
Github repo: https://github.com/minnesotanlp/Quantifying-Annotation-Disagreement
Source Data: Scruples-dilemmas (Lourie, Bras, and Choi 2021)
multi-step-moral-dilemmas
[!NOTE]
This is a copy from: https://isir-wuya.github.io/Multi-step-Moral-Dilemmas/
Paper:
@misc{wu2025staircaseethicsprobingllm,
title={The Staircase of Ethics: Probing LLM Value Priorities through Multi-Step Induction to Complex Moral Dilemmas},
author={Ya Wu and Qiang Sheng and Danding Wang and Guang Yang and Yifan Sun and Zhengjia Wang and Yuyan Bu and Juan Cao},
year={2025},
eprint={2505.18154},
archivePrefix={arXiv},
primaryClass={cs.CL}… See the full description on the dataset page: https://huggingface.co/datasets/DebateLabKIT/multi-step-moral-dilemmas.laurashin_The_Chopping_Block_The_EVM_Parallelization_Dilemma_Solanas_Network_Congestion_and_Avi_Tessera-WADT-Dilemmas
Tessera WADT Dilemmas
WADT — Wike Adversarial Dilemma Training. 658 structured ethical-dilemma pairs
built to train a model to commit to a decision under pressure instead of hedging,
flattering, or deferring — the opposite instinct of a sycophantic model, applied
to hard cases with no clean answer.
Why this exists
Most "AI ethics" training data teaches a model to discuss dilemmas. WADT trains
a model to decide — every example follows a fixed structure: name the… See the full description on the dataset page: https://huggingface.co/datasets/AIIT-Threshold/Tessera-WADT-Dilemmas.prisoners_dilemma_dpo_phihan-moral-dilemma-scenarios-v1
Humanoid Moral Dilemma Scenarios
A curated dataset of ethical and moral dilemmas
designed to train humanoid AI on value-aware decision-making.
Use Cases
Ethical reasoning
Safety alignment
Value-sensitive AI behavior
Fields
dilemma_context
conflicting_values
recommended_constraint
acceptable_actions
Part of
Humanoid Network (HAN)
License
MIT
beyond_one_world-dilemma@misc{2510.14351,
Author = {Perapard Ngokpol and Kun Kerdthaisong and Pasin Buakhaw and Pitikorn Khlaisamniang and Supasate Vorathammathorn and Piyalitt Ittichaiwong and Nutchanon Yongsatianchot},
Title = {Beyond One World: Benchmarking Super Heros in Role-Playing Across Multiversal Contexts},
Year = {2025},
Eprint = {arXiv:2510.14351},
}
daily_dilemmas
DailyDilemmas - Revealing Value Preferences of LLMs with Quandaries of Daily Life
Link: Paper
Description of DailyDilemma
DailyDilemma is a dataset of 1,360 moral dilemmas encountered in everyday life. Each dilemma includes two possible actions and with each action, the affected parties and human values invoked.
We evaluated LLMs on these dilemmas to determine what action they will take and the values represented by these actions
Dataset details… See the full description on the dataset page: https://huggingface.co/datasets/hoshoic/daily_dilemmas.daily_dilemmas-self
daily_dilemmas-self
subset of kellycyy/daily_dilemmas, converted into symmetric per-value labels.
Preprocessing
This dataset applies four simple preprocessing steps:
keep only raw rows where party == "You" exactly
remove duplicate value mentions that appear on both to_do and not_to_do for the same dilemma
symmetrize the remaining value tags across to_do and not_to_do
keep only dilemma pairs with at least one remaining explicit-You value, then sort pairs by label count… See the full description on the dataset page: https://huggingface.co/datasets/wassname/daily_dilemmas-self.ethical-dilemmas-dataset
ethical-dilemmas-dataset
Dataset Description
The Ethical Dilemmas Dataset serves as a resource for exploring and analyzing various ethical scenarios that challenge individuals and organizations. This dataset is designed to facilitate discussions on moral reasoning and decision-making processes across diverse contexts. Each entry presents a structured ethical dilemma, accompanied by a Chain-of-Thought (CoT) analysis that outlines key considerations and potential… See the full description on the dataset page: https://huggingface.co/datasets/Mobiusi/ethical-dilemmas-dataset.adaption-ethical-dilemma-scenarios
This dataset is a remastered version prepared using Adaption's Adaptive Data platform.
adaption-ethical_dilemma_scenarios
This dataset contains a collection of complex ethical dilemmas presented as multiple-choice questions, focusing on military command decisions, professional integrity, and moral conflicts. Each sample provides a detailed scenario involving high-stakes consequences, such as civilian casualties, war crimes, or financial exploitation, followed by five potential… See the full description on the dataset page: https://huggingface.co/datasets/webdevsha/adaption-ethical-dilemma-scenarios.adaption_complex_clinical_dilemmas_v1gptoss_dilemma_choices
Ethical Conflict Simulation — gpt-oss-20b Traces
A dataset of 45 ethical-dilemma prompt–response pairs (15 scenarios × 3 independent runs) collected from OpenAI's open-weight model gpt-oss-20b. Each row contains the full prompt, the model's complete response, its exposed chain-of-thought reasoning trace, and the wall-clock response time in milliseconds. The goal was to map where the model's moral decision-making is consistent, where it drifts, and what contextual triggers cause… See the full description on the dataset page: https://huggingface.co/datasets/sems/gptoss_dilemma_choices.personalisation-daily-dilemmasprisoners_dilemmaneuro_diag_dilemma_v1cardio_diag_dilemma_v2trolley-dilemmadebugging-failure-dilemma-v10diag_dilemma_v2endo_diag_dilemma_v3