MUCnoHARM@GermEval 2026共享任务:基于检索的上下文学习用于诽谤罪及其不足之处
MUCnoHARM@GermEval Shared Task 2026: Retrieval-based In-Context Learning for Defamatory Offences, and Where It Falls Short
浏览论文内容
中文总结 AI 辅助
本研究评估多种基于检索的上下文学习策略检测德国刑法第185-187条诽谤罪,发现少样本优于零样本,但检索方法增益有限,模型选择影响最大,模型适合分诊而非自主审核。
中文摘要 AI 辅助
鉴于仇恨言论在网上无处不在,自动检测至关重要,尤其是对于具有刑事相关性的社交媒体帖子。我们研究了多种基于检索的上下文学习(RetICL)策略,用于检测《德国刑法典》第185至187条下的诽谤罪(这是GermEval 2026子任务4的主题)。少样本提示优于零样本,但基于检索的方法相比随机示例仅带来边际提升,甚至落后于优化后的静态示例集。提供具体的法律知识有所帮助,但模型选择的影响超过所有其他系统选择。模型过度预测刑事相关性,同时仍遗漏26%至57%的刑事相关帖子,使其适合分诊而非自主审核。
英文摘要
With hate speech being ubiquitous online, automatic detection is crucial, in particular when it comes to criminally relevant social media posts. We study a variety of retrieval-based in-context learning (RetICL) strategies for detecting defamatory offences under §§ 185-187 StGB (the subject of GermEval 2026 Subtask 4). Few-shot prompting beats zero-shot, but retrieval-based approaches offer only marginal gains over random demonstrations, and even fall behind an optimised static set of demonstrations. Providing concrete legal knowledge helps, yet model choice outweighs every other system choice. Models over-predict criminal relevance while still missing 26-57% of criminally relevant posts, suiting them for triage rather than autonomous moderation.
发表机构
- Central Office for Information Technology in the Security Sector (ZITiS)(安全领域信息技术中央办公室(ZITiS))
- LMU Munich(慕尼黑大学)
- Munich Center for Machine Learning (MCML)(慕尼黑机器学习中心)
机构由 AI 辅助整理,请以论文原文为准。