主动请求：用于预测在线内容审核中规则违规的问答方法
Asking For It: Question-Answering for Predicting Rule Infractions in Online Content Moderation
AI总结：
本文提出ModQ问答框架，利用社区规则进行内容审核，通过抽取式和选择题问答变体在Reddit和Lemmy数据上训练，优于基线并泛化至新社区。
摘要：
在线社区依赖平台政策与社区自定规则的结合来界定可接受行为并维持秩序。然而，这些规则在不同社区间差异巨大，随时间演变，且执行不一致，给透明度、治理和自动化带来挑战。在本文中，我们对规则与其执行之间的关系进行大规模建模，引入ModQ，一种新颖的面向规则敏感内容审核的问答框架。与以往基于分类或生成的方法不同，ModQ在推理时以全部社区规则为条件，并识别出最适用于给定评论的规则。我们实现了两种模型变体——抽取式问答和多选题问答——并在来自Reddit和Lemmy的大规模数据集上训练它们，其中Lemmy数据集由我们根据公开的审核日志和规则描述构建。两种模型在识别与审核相关的规则违规方面均优于最先进的基线，同时保持轻量级和可解释性。值得注意的是，ModQ模型能有效泛化到未见过的社区和规则，支持低资源审核设置和动态治理环境。
英文摘要：
Online communities rely on a mix of platform policies and community-authored rules to define acceptable behavior and maintain order. However, these rules vary widely across communities, evolve over time, and are enforced inconsistently, posing challenges for transparency, governance, and automation. In this paper, we model the relationship between rules and their enforcement at scale, introducing ModQ, a novel question-answering framework for rule-sensitive content moderation. Unlike prior classification or generation-based approaches, ModQ conditions on the full set of community rules at inference time and identifies which rule best applies to a given comment. We implement two model variants - extractive and multiple-choice QA - and train them on large-scale datasets from Reddit and Lemmy, the latter of which we construct from publicly available moderation logs and rule descriptions. Both models outperform state-of-the-art baselines in identifying moderation-relevant rule violations, while remaining lightweight and interpretable. Notably, ModQ models generalize effectively to unseen communities and rules, supporting low-resource moderation settings and dynamic governance environments.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
