Reproduce “A Theory of Human-Like Few-Shot Learning” 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
Zhiying Jiang, Rui Wang, Dongbo Bu, Ming Li
The authors present few-shot classification as comparing test items with labeled examples by symmetric compression distance, approximated with a variational autoencoder for images and gzip for text. Across five 5-shot image tests, their method reports the highest accuracy among listed methods; on CIFAR-10 it scores 35.3±2.9 versus 22.2±1.6 for a latent-representation control. These results cover benchmarks, not human cognition.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.