anonymous-review-anonymous/HarmMetric_Eval

HarmMetric Eval Some of the contents of the dataset may be offensive to some readers. This is the official repository of HarmMetric Eval: Benchmarking Metrics and Judges for LLM Harmfulness Assessment. This repository contains three datasets: the benchmark dataset dataset.jsonl, the training data for HarmClassifier train_data.jsonl, and the test dataset test_data.jsonl. Due to the limitation of the Hugging Face Dataset Viewer in displaying data with structures that differ from… See the full description on the dataset page: https://huggingface.co/datasets/anonymous-review-anonymous/HarmMetric_Eval.