利用思维链监督的强化学习进行仇恨和宣传模因的可解释检测
Adapting Reinforcement Learning with Chain-of-Thought Supervision for Explainable Detection of Hateful and Propagandistic Memes
AI总结：
提出基于强化学习的后训练方法，结合任务特定奖励和组相对策略优化（GRPO），提升思考型多模态大语言模型在仇恨和宣传模因检测中的分类性能和解释质量。
摘要：
仇恨和宣传模因利用图像与文本之间的相互作用来传达有害意图，而这两种模态单独都无法揭示这种意图。尽管基于思考的多模态大语言模型（MLLMs）在视觉-语言理解方面取得了进展，但它们在模因内容审核中的应用仍未得到充分探索。我们提出了一种基于强化学习的后训练方法，通过任务特定奖励和组相对策略优化（GRPO）来提高思考型MLLMs的分类性能和基于参考的解释质量。具体来说，我们（i）对现成的MLLMs在英语和阿拉伯语基准上的仇恨和宣传模因理解进行了系统的实证研究，（ii）通过蒸馏和多LLM细粒度宣传标注，用弱监督的思维链（CoT）理由扩展了现有的模因数据集，（iii）引入了一个基于GRPO的目标函数，带有思考长度正则化，联合优化分类准确性和解释质量，以及（iv）研究基于共识伪标签的无标签模因的自监督GRPO。在Hateful Memes和ArMeme基准上的实验表明，我们的方法在FHM准确率（从79.9%提高到82.0%，提升高达2.1%）和ArMeme宏F1（从0.536提高到0.612，提升高达7.6个百分点，附带解释；与原始ArMeme基准相比提升6.1个百分点）上优于先前报告的结果，同时生成自然语言解释。在ArMeme上，序列分类基线在原始准确率方面仍然更强，而我们的方法提供了更平衡的每类性能以及解释。我们公开发布了代码、数据扩展和评估资源。
英文摘要：
Hateful and propagandistic memes exploit the interplay between images and text to convey harmful intent that neither modality reveals alone. Although thinking-based multimodal large language models (MLLMs) have advanced vision-language understanding, their application to meme content moderation remains underexplored. We propose a reinforcement learning-based post-training method that improves classification performance and reference-based explanation quality in thinking-based MLLMs via task-specific rewards and Group Relative Policy Optimization (GRPO). Concretely, we (i) conduct a systematic empirical study of off-the-shelf MLLMs for hateful and propagandistic meme understanding across English and Arabic benchmarks, (ii) extend existing meme datasets with weakly supervised chain-of-thought (CoT) rationales via distillation and multi-LLM fine-grained propaganda annotations, (iii) introduce a GRPO-based objective with thinking-length regularization that jointly optimizes classification accuracy and explanation quality, and (iv) investigate self-supervised GRPO on unlabeled memes using consensus-based pseudo-labels. Experiments on the Hateful Memes and ArMeme benchmarks show that our approach improves over previously reported results on FHM accuracy (up to +2.1%, from 79.9% to 82.0%) and on ArMeme macro-F1 (up to +7.6 points, from 0.536 to 0.612 with explanations; +6.1 compared to the original ArMeme benchmark), while also generating natural-language explanations. On ArMeme, sequence-classification baselines remain stronger in terms of raw accuracy, whereas our approach provides more balanced per-class performance along with explanations. We publicly release our code, data extensions, and evaluation resources.
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。
