HateMirage: An Explainable Multi-Dimensional Dataset for Decoding Faux Hate and Subtle Online Abuse

A research team from IIIT Hyderabad and IIT Delhi has introduced HateMirage, a groundbreaking dataset designed to decode the complex relationship between misinformation and subtle online hate speech. Accepted at LREC 2026, this work addresses a critical gap in online safety research by focusing on 'faux hate'—harmful intent embedded within misleading or manipulative narratives that existing datasets often miss. The researchers constructed HateMirage by identifying widely debunked misinformation claims from fact-checking sources and tracing related YouTube discussions, creating a corpus that moves beyond overt toxicity to capture nuanced social harm.

HateMirage contains 4,530 user comments, each annotated along three interpretable dimensions: Target (who is affected), Intent (the underlying motivation), and Implication (potential social impact). This multi-dimensional explanation framework surpasses previous datasets like HateXplain and HARE, which offered only token-level or single-dimensional reasoning. The team benchmarked multiple open-source language models using ROUGE-L F1 and Sentence-BERT similarity metrics, revealing that explanation quality depends more on pretraining diversity and reasoning-oriented data than on model scale alone. By coupling misinformation reasoning with harm attribution, HateMirage establishes a new benchmark for developing more interpretable and responsible AI systems for content moderation.