CoolFace
20 results

moral

demelin /moral_storiesMoral Stories is a crowd-sourced dataset of structured, branching narratives for the study of grounded, goal-oriented social reasoning. For detailed information, see https://aclanthology.org/2021.emnlp-main.54.pdf.textmultiple-choice100K<n<1M34 likes4.9k downloads4y agoHugging Faceninoscherrer /moralchoice Dataset Card for MoralChoice Homepage: Coming Soon Paper: Coming soon Repository: https://github.com/ninodimontalcino/moralchoice Point of Contact: Nino Scherrer & Claudia Shi Dataset Summary MoralChoice is a survey dataset to evaluate the moral beliefs encoded in LLMs. The dataset consists of: Survey Question Meta-Data: 1767 hypothetical moral scenarios where each scenario consists of a description / context and two potential actions Low-Ambiguity Moral Scenarios… See the full description on the dataset page: https://huggingface.co/datasets/ninoscherrer/moralchoice.textn<1K6 likes2.2k downloads3y agoHugging FaceAmbarishGK /MoRAL-nuscenes-v3image0 likes576 downloads6mo agoHugging FaceAmbarishGK /MoRAL-nuscenesimage0 likes433 downloads6mo agoHugging Faceagentic-moral-alignment /mtma mtma — run artifacts Artifacts from the Multi-Turn Moral Alignment project. records/structure3/ — MACHIAVELLI env-structure rollouts (v3): 20,123 start points, 2.57M random-policy episodes over a game x depth x walk-seed grid. See records/structure3/README.md for the layout, record schema and how to extract. records/logs/ — collection logs for the above. runs/ — training-run artifacts (per-run logs, metrics, rollout dumps). records/ is analysis data; runs/ is training output.… See the full description on the dataset page: https://huggingface.co/datasets/agentic-moral-alignment/mtma.1 likes407 downloads25d agoHugging Facelocuslab /moral_education Moral Education Dataset Overview The Moral Education dataset is a collection of instructional content derived from potentially harmful web text that has been transformed into educational material. This dataset is part of the safety pretraining initiative described in the SafeLM research paper, which aims to develop natively safe language models by embedding safety directly into the pretraining process. Dataset Creation The source data comes from RefuseWeb… See the full description on the dataset page: https://huggingface.co/datasets/locuslab/moral_education.text1M<n<10M2 likes391 downloads4mo agoHugging Face