SciEducator: Scientific Video Understanding and Educating via Deming-Cycle Multi-Agent System

2025-11-22

AI Summary

Current multimodal large language models (MLLMs) and video agents struggle to integrate external domain expertise and perform rigorous multi-step reasoning for scientific video understanding and education. To address this, we propose the first PDCA (Plan-Do-Check-Act)-driven multi-agent framework for scientific video understanding, establishing an iterative, self-evolving cognitive architecture that synergistically integrates MLLMs, video agents, and domain-expert-knowledge-guided reasoning. Our method enables cross-modal (text, visual, audio) pedagogical content generation and is systematically evaluated on SciVBench, a novel, in-house scientific video benchmark. Experiments demonstrate substantial gains over state-of-the-art models—including Gemini and GPT-4o—and advanced video agents. This work pioneers the integration of the PDCA cycle into AI-driven educational reasoning, setting a new standard for scientific video understanding and intelligent instructional systems.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Computer Vision: Video Understanding & Activity Analysis
• Cognitive Modeling & Cognitive Systems: Agent Architectures

Application Category

• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Search and Retrieval-Augmented AI: Agentic search
• Responsible Web: Machine-in-the-loop, human agency and autonomy

Abstract

Recent advancements in multimodal large language models (MLLMs) and video agent systems have significantly improved general video understanding. However, when applied to scientific video understanding and educating, a domain that demands external professional knowledge integration and rigorous step-wise reasoning, existing approaches often struggle. To bridge this gap, we propose SciEducator, the first iterative self-evolving multi-agent system for scientific video comprehension and education. Rooted in the classical Deming Cycle from management science, our design reformulates its Plan-Do-Study-Act philosophy into a self-evolving reasoning and feedback mechanism, which facilitates the interpretation of intricate scientific activities in videos. Moreover, SciEducator can produce multimodal educational content tailored to specific scientific processes, including textual instructions, visual guides, audio narrations, and interactive references. To support evaluation, we construct SciVBench, a benchmark consisting of 500 expert-verified and literature-grounded science QA pairs across five categories, covering physical, chemical, and everyday phenomena. Extensive experiments demonstrate that SciEducator substantially outperforms leading closed-source MLLMs (e.g., Gemini, GPT-4o) and state-of-the-art video agents on the benchmark, establishing a new paradigm for the community.

Problem

• Existing video AI struggles with scientific content understanding
• Professional knowledge integration and step-wise reasoning are challenging
• Lack of tailored educational content for scientific processes

Innovation

• Multi-agent system using Deming Cycle for video understanding
• Self-evolving reasoning mechanism for scientific activities
• Generates multimodal educational content with tailored instructions