arXiv:2606.07433v1 Announce Type: new 
Abstract: Video understanding is being rapidly transformed by multimodal large language models (MLLMs), as research moves from short clips to long, multimodal, and knowledge-intensive video scenarios. These scenarios require models to handle sparse evidence, long-range dependencies, multimodal alignment, and reliable inference under limited computational budgets. This work presents a human-view perspective on LLM-based video understanding, organized around three functional abilities: watching, remembering, and reasoning. Rather than treating video tasks as isolated benchmarks, this view provides a unified structure for analyzing how video MLLMs acquire evidence, preserve context, and produce grounded outputs. We introduce a formulation that characterizes video understanding systems by their perceptual representations, memory states, reasoning traces, and final predictions. Based on this formulation, we identify challenges in spatio-temporal perception, efficient long-video processing, memory modeling, streaming understanding, and faithful reasoning. Representative methods are organized by their roles in video MLLM systems. Watching covers fine-grained, comprehensive, audio-visual, and efficient perception. Remembering includes offline and streaming memory, while reasoning covers text-only reasoning and thinking with videos. We further examine application domains such as egocentric, sports, instructional, medical, and narrative videos, and cover training datasets and evaluation benchmarks across task types, supervision formats, modalities, and capability dimensions. Finally, we outline open problems and future directions for scalable, memory-aware, and evidence-grounded video intelligence. Related works will be continuously traced at https://github.com/marinero4972/Awesome-HumanView-VideoUnderstanding.

تشهد الفهم الفيديوي تقدمًا سريعًا بفضل نماذج اللغة متعددة الوسائط (MLLMs)، حيث تنتقل الأبحاث من مقاطع الفيديو القصيرة إلى سيناريوهات فيديو طويلة ومعقدة تتطلب التعامل مع أدلة متفرقة وارتباطات طويلة المدى. يركز هذا العمل على ثلاث قدرات وظيفية: المشاهدة، والتذكر، والتفكير، مما يوفر هيكلًا موحدًا لتحليل كيفية اكتساب نماذج MLLM للأدلة، والحفاظ على السياق، وإنتاج مخرجات موثوقة.

Los avances recientes en la comprensión de video están siendo impulsados por modelos de lenguaje multimodal (MLLM), que están evolucionando de analizar clips cortos a abordar escenarios de video largos y complejos que requieren manejar evidencia escasa y dependencias a largo plazo. Este nuevo enfoque enfatiza tres habilidades funcionales: observar, recordar y razonar, proporcionando un marco estructurado para analizar cómo los MLLM procesan datos de video.

Des avancées récentes dans la compréhension vidéo sont impulsées par des modèles de langage multimodaux (MLLM), qui évoluent de l'analyse de courts extraits à la gestion de scénarios vidéo longs et complexes nécessitant le traitement de preuves éparses et de dépendances à long terme. Cette nouvelle approche met l'accent sur trois capacités fonctionnelles : regarder, se souvenir et raisonner, fournissant un cadre structuré pour analyser comment les MLLM traitent les données vidéo.

Recent advancements in video understanding are being driven by multimodal large language models (MLLMs), which are evolving from analyzing short clips to tackling long, complex video scenarios that require handling sparse evidence and long-range dependencies. This new approach emphasizes three functional abilities: watching, remembering, and reasoning, providing a structured framework for analyzing how MLLMs process video data.