# Conditioning Large Language Models on Legal Systems? Detecting Punishable Hate Speech

arXiv: https://arxiv.org/abs/2506.03009 · HTML: https://mlsanity.com/paper/2506.03009/

## Key facts

- Cited by 2 papers in the map (data through 14 Oct 2026).
- Builds on 17 of its 33 references that are in the map.
- First posted on arXiv 3 Jun 2025.

## Builds on (most cited first, 17 of 17)

- [BART: Denoising Sequence-to-Sequence Pre-training for Natural Language Generation, Translation, and Comprehension](https://mlsanity.com/paper/1910.13461/) (2019)
- [Lost in the Middle: How Language Models Use Long Contexts](https://mlsanity.com/paper/2307.03172/) (2023)
- [Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone](https://mlsanity.com/paper/2404.14219/) (2024)
- [Quantifying Language Models' Sensitivity to Spurious Features in Prompt Design or: How I learned to start worrying about prompt formatting](https://mlsanity.com/paper/2310.11324/) (2023)
- [A Survey on In-context Learning](https://mlsanity.com/paper/2301.00234/) (2022)
- [Neural Legal Judgment Prediction in English](https://mlsanity.com/paper/1906.02059/) (2019)
- [Let Me Speak Freely? A Study on the Impact of Format Restrictions on Performance of Large Language Models](https://mlsanity.com/paper/2408.02442/) (2024)
- [Detecting Hate Speech with GPT-3](https://mlsanity.com/paper/2103.12407/) (2021)
- [Legal Prompt Engineering for Multilingual Legal Judgement Prediction](https://mlsanity.com/paper/2212.02199/) (2022)
- [Analyzing the Targets of Hate in Online Social Media](https://mlsanity.com/paper/1603.07709/) (2016)
- [You Only Prompt Once: On the Capabilities of Prompt Learning on Large Language Models to Tackle Toxic Content](https://mlsanity.com/paper/2308.05596/) (2023)
- [Legal Syllogism Prompting: Teaching Large Language Models for Legal Judgment Prediction](https://mlsanity.com/paper/2307.08321/) (2023)
- [Sensitivity and Robustness of Large Language Models to Prompt Template in Japanese Text Classification Tasks](https://mlsanity.com/paper/2305.08714/) (2023)
- [ToKen: Task Decomposition and Knowledge Infusion for Few-Shot Hate Speech Detection](https://mlsanity.com/paper/2205.12495/) (2022)
- [A Legal Approach to Hate Speech: Operationalizing the EU's Legal Framework against the Expression of Hatred as an NLP Task](https://mlsanity.com/paper/2004.03422/) (2020)
- [The Factuality of Large Language Models in the Legal Domain](https://mlsanity.com/paper/2409.11798/) (2024)
- [HumBEL: A Human-in-the-Loop Approach for Evaluating Demographic Factors of Language Models in Human-Machine Conversations](https://mlsanity.com/paper/2305.14195/) (2023)

## Built on by (most cited first, 2 of 2)

- [Beyond Imperfect Alternatives with Rulemapping: A Neuro-Symbolic Case Study on Online Hate Speech](https://mlsanity.com/paper/2605.16280/) (2026)
- [MUCnoHARM@GermEval Shared Task 2026: Retrieval-based In-Context Learning for Defamatory Offences, and Where It Falls Short](https://mlsanity.com/paper/2609.09791/) (2026)

## Abstract (arXiv)

The assessment of legal problems requires the consideration of a specific legal system and its levels of abstraction, from constitutional law to statutory law to case law. The extent to which Large Language Models (LLMs) internalize such legal systems is unknown. In this paper, we propose and investigate different approaches to condition LLMs at different levels of abstraction in legal systems. This paper examines different approaches to conditioning LLMs at multiple levels of abstraction in legal systems to detect potentially punishable hate speech. We focus on the task of classifying whether a specific social media posts falls under the criminal offense of incitement to hatred as prescribed by the German Criminal Code. The results show that there is still a significant performance gap between models and legal experts in the legal assessment of hate speech, regardless of the level of abstraction with which the models were conditioned. Our analysis revealed, that models conditioned on abstract legal knowledge lacked deep task understanding, often contradicting themselves and hallucinating answers, while models using concrete legal knowledge performed reasonably well in identifying relevant target groups, but struggled with classifying target conducts.