arXiv:2602.13748v2 Announce Type: replace-cross 
Abstract: Multimedia Event Extraction (MEE) aims to identify events and their arguments from documents that contain both text and images. It requires grounding event semantics across different modalities. Progress in MEE is limited by the lack of annotated training data. M2E2 is the only established benchmark, but it provides annotations only for evaluation. This makes direct supervised training impractical. Existing methods mainly rely on cross-modal alignment or inference-time prompting with Vision--Language Models (VLMs). These approaches do not explicitly learn structured event representations and often produce weak argument grounding in multimodal settings. To address these limitations, we propose RMPL, a Relation-aware Multi-task Progressive Learning framework for MEE under low-resource conditions. RMPL incorporates heterogeneous supervision from unimodal event extraction and multimedia relation extraction with stage-wise training. The model is first trained with a unified schema to learn shared event-centric representations across modalities. It is then fine-tuned for event mention identification and argument role extraction using mixed textual and visual data. Experiments on the M2E2 benchmark with multiple VLMs show consistent improvements across different modality settings.

تم اقتراح إطار عمل جديد يسمى RMPL، أو التعلم التقدمي متعدد المهام القائم على العلاقات، لتحسين استخراج الأحداث متعددة الوسائط (MEE)، الذي يهدف إلى تحديد الأحداث وحججها من الوثائق التي تحتوي على نصوص وصور. يعالج هذا التطور القيود التي تواجهها الأساليب الحالية التي تكافح في ظروف الموارد المنخفضة ونقص بيانات التدريب المعلّمة، خاصة في سياق معيار M2E2.

Se ha propuesto un nuevo marco llamado RMPL, o Aprendizaje Progresivo Multi-tarea Sensible a Relaciones, para mejorar la Extracción de Eventos Multimedia (MEE), que identifica eventos y sus argumentos en documentos que contienen tanto texto como imágenes. Este desarrollo aborda las limitaciones de los métodos existentes que luchan en condiciones de bajos recursos y la falta de datos de entrenamiento anotados, especialmente en el contexto del referente M2E2.

Un nouveau cadre appelé RMPL, ou Apprentissage Progressif Multi-tâches Sensible aux Relations, a été proposé pour améliorer l'Extraction d'Événements Multimédias (MEE), qui identifie les événements et leurs arguments dans des documents contenant à la fois du texte et des images. Ce développement répond aux limitations des méthodes existantes qui peinent dans des conditions de faible ressource et au manque de données d'entraînement annotées, en particulier dans le contexte de la référence M2E2.

A new framework called RMPL, or Relation-aware Multi-task Progressive Learning, has been proposed to enhance Multimedia Event Extraction (MEE), which identifies events and their arguments from documents containing both text and images. This development addresses the limitations of existing methods that struggle with low-resource conditions and the lack of annotated training data, particularly in the context of the M2E2 benchmark.