LIBERTy-Workplace Violence is one of the three datasets released as part of theLIBERTy (LLM-based Interventional Benchmark for Explainability with Real Targets) benchmark.
LIBERTy is designed to evaluate concept-based explanation methods in NLP
under a causal and counterfactual framework.Each dataset exposes structured dependencies between high-level semantic concepts
and model predictions, enabling quantitative evaluation of… See the full description on the dataset page: https://huggingface.co/datasets/GilatToker/Liberty-Violence.
