# ThreatCore: A Benchmark for Explicit and Implicit Threat Detection

arXiv: https://arxiv.org/abs/2605.10563 · HTML: https://mlsanity.com/paper/2605.10563/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 7 of its 21 references that are in the map.
- First posted on arXiv 11 May 2026.

## Builds on (most cited first, 7 of 7)

- [Predicting the Type and Target of Offensive Posts in Social Media](https://mlsanity.com/paper/1902.09666/) (2019)
- [Learning from the Worst: Dynamically Generated Datasets to Improve Online Hate Detection](https://mlsanity.com/paper/2012.15761/) (2020)
- [Latent Hatred: A Benchmark for Understanding Implicit Hate Speech](https://mlsanity.com/paper/2109.05322/) (2021)
- [ETHOS: an Online Hate Speech Detection Dataset](https://mlsanity.com/paper/2006.08328/) (2020)
- [Hate Speech Detection: A Solved Problem? The Challenging Case of Long Tail on Twitter](https://mlsanity.com/paper/1803.03662/) (2018)
- [Mapping the Italian Telegram Ecosystem: Communities, Toxicity, and Hate Speech](https://mlsanity.com/paper/2504.19594/) (2025)
- [AMAQA: A Metadata-based QA Dataset for RAG Systems](https://mlsanity.com/paper/2505.13557/) (2025)

## Abstract (arXiv)

Threat detection in Natural Language Processing lacks consistent definitions and standardized benchmarks, and is often conflated with broader phenomena such as toxicity, hate speech, or offensive language. In this work, we introduce ThreatCore, a public available benchmark dataset for fine-grained threat detection that distinguishes between explicit threats, implicit threats, and non-threats. The dataset is constructed by aggregating multiple publicly available resources and systematically re-annotating them under a unified operational definition of threat, revealing substantial inconsistencies across existing labels. To improve the coverage of underrepresented cases, particularly implicit threats, we further augment the dataset with synthetic examples, which are manually validated using the same annotation protocol adopted for the re-annotation of the public datasets, ensuring consistency across all data sources. We evaluate Perspective API, zero-shot classifiers, and recent language models on ThreatCore, showing that implicit threats remain substantially harder to detect than explicit ones. Our results also indicate that incorporating Semantic Role Labeling as an intermediate representation can improve performance by making the structure of harmful intent more explicit. Overall, ThreatCore provides a more consistent benchmark for studying fine-grained threat detection and highlights the challenges that current models still face in identifying indirect expressions of harmful intent.