Reproduce “Model Ratatouille: Recycling Diverse Models for Out-of-Distribution Generalization” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Alexandre Ramé, Kartik Ahuja, Jianyu Zhang, Matthieu Cord, Léon Bottou, David Lopez-Paz
The authors present ratatouille, a method for improving out-of-distribution image classification by reusing diverse fine-tunings of a shared pretrained model. It fine-tunes separate auxiliary-initialized models on the target task, then averages their weights. On DomainBed, uniform ratatouille reaches 68.1% accuracy, versus 67.6% for model soups. This reported average covers five datasets and does not establish improvement across all domains.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.