SoccerRef-Agents: Multi-Agent AI System for Automated Soccer Refereeing
A new multi-agent AI framework called SoccerRef-Agents aims to automate soccer refereeing with explainable decisions. The system integrates a multimodal benchmark, SoccerRefBench, containing over 1,200 referee theory questions and 600 foul video clips. It also uses a vector-based knowledge base, RefKnowledgeDB, built from the latest 'Laws of the Game' and classic case databases. The architecture employs cross-modal retrieval-augmented generation (RAG) to bridge visual and semantic gaps. This research addresses the limitation of existing AI approaches that focus on isolated video perception without reasoning about foul scenarios. The work is published on arXiv under ID 2604.23392.
Key facts
- SoccerRef-Agents is a multi-agent decision-making framework for soccer refereeing.
- It uses a multimodal benchmark SoccerRefBench with over 1,200 referee theory questions and 600 foul video clips.
- RefKnowledgeDB is a vector-based knowledge base using the latest 'Laws of the Game' and a classic case database.
- The multi-agent architecture collaborates via cross-modal RAG to bridge visual and semantic gaps.
- Current AI-assisted approaches in soccer refereeing lack understanding and reasoning of foul scenarios.
- The research is published on arXiv with ID 2604.23392.
- The system aims for fair, accurate, and explainable decisions.
- The paper is from arXiv, a preprint repository.
Entities
Institutions
- arXiv