发表机构
Graphic Era University（葛拉蒂克大学）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
该研究提出基于结构化论证辩论的多智能体偏见检测框架MABPD，无需特定任务训练，在BABE基准上接近监督SOTA，跨数据集也具迁移性，可有效检测新闻媒体偏见。
AI 中文摘要
新闻文章中的媒体偏见通过微妙的语言线索运作，如带有倾向性的语言、选择性框架和战略性省略，这些线索难以被单一模型检测，传统上需要大型标注语料库进行监督训练。本文探究结构化多智能体审议能否作为该任务的无训练替代方案，取代监督分类。我们提出MABPD（Multi-Agent Bias Probing & Detection，多智能体偏见探测与检测），该流程包含三个专门的大语言模型（LLM）智能体，从互补视角分析文章，并通过结构化论证辩论（Structured Argument Debate，SAD）协议解决分歧。SAD采用受领域驱动的不对称举证责任——无文本证据支持的偏见主张权重为零，结合角色加权投票与共识后验证，以明确的审议结构取代特定任务的监督决策边界。消融实验证实，驱动性能的是这种结构化审议而非单纯的智能体并行：移除辩论模块会使F1值最多降低10.6个百分点。在BABE基准（4121条专家标注句子）上，MABPD在保留测试集上的宏观F1值达83.4%，与监督式SOTA（MAGPIE，宏观F1值84.1%；Horych等人，2024）仅相差0.7个百分点，且无需对标注数据进行任何特定任务训练或阈值调整。在SemEval 2019 HyperPartisan语料库（644篇文章）上的跨数据集评估显示，零样本准确率达75.0%，与监督式SOTA准确率（82.2%；Kiesel等人，2019）相差7.2个百分点，证实其可跨标注机制迁移。我们发布了完整流程与评估代码。
英文摘要
Media bias in news articles operates through subtle linguistic cues---loaded language, selective framing, and strategic omission---that resist single-model detection and have traditionally required large annotated corpora for supervised training. We ask whether structured multi-agent deliberation can serve as a principled, training-free alternative to supervised classification for this task. We introduce MABPD (Multi-Agent Bias Probing & Detection), a pipeline in which three specialized LLM agents analyze an article from complementary perspectives and resolve disagreements through a Structured Argument Debate (SAD) protocol. SAD implements a domain-motivated asymmetric burden of proof---biased claims without grounded textual evidence carry zero weight---combined with role-weighted voting and post-consensus verification, replacing task-specific supervised decision boundaries with explicit deliberative structure. Ablation confirms that this structured deliberation, not mere agent parallelism, drives performance: removing the debate module reduces F1 by up to 10.6 points. On the BABE benchmark (4,121 expert-annotated sentences), MABPD achieves 83.4% macro F1 on the held-out test split---within 0.7 percentage points (pp) of the supervised SOTA (MAGPIE, 84.1% macro F1; Horych et al., 2024)---without any task-specific training or threshold tuning on annotated data. Cross-dataset evaluation on the SemEval 2019 HyperPartisan corpus (644 articles) yields 75.0% zero-shot accuracy, within 7.2 pp of the supervised SOTA accuracy (82.2%; Kiesel et al. 2019), confirming transfer across annotation regimes. We release the full pipeline and evaluation code.
Comments20 pages, 6 figures. Accepted to the EMNLP 2026 Main Conference. Code: https://github.com/Subaru-5999/MABPD
