A benchmark dataset for evaluating moral decision-making of foundation models from an agent-centric perspective.
ABHBench provides a collection of interactive moral dilemma scenarios extracted from narrative-driven video games. Each scenario presents an agent with contextual information and requires choosing between morally significant actions.
The dataset contains 7 chapters (dataset_flow_01 through dataset_flow_27), each… See the full description on the dataset page:
https://huggingface.co/datasets/abhbench/ABHBench.
