Modelpedia: A Catalog of Model Findings for the Meta-Science of AI
Modelpedia turns scattered model-behavior claims into a searchable catalog for studying how AI models are investigated.
The paper describes an automated, LLM-assisted framework that extracts findings from published work and links them to the relevant model, dataset, method, and concept. Its prototype was applied to accepted ICLR 2024 and 2025 papers, producing more than a thousand findings. The authors also use the catalog for a meta-analysis of research practice around AI models. ArXiv · AI/CL/LG's note
The paper describes an automated, LLM-assisted framework that extracts findings from published work and links them to the relevant model, dataset, method, and concept. Its prototype was applied to accepted ICLR 2024 and 2025 papers, producing more than a thousand findings. The authors also use the catalog for a meta-analysis of research practice around AI models. ArXiv · AI/CL/LG's note
score 4