Reproduce “Robust Meta-Representation Learning via Global Label Inference and Classification” 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
Ruohan Wang, Isak Falk, Massimiliano Pontil, Carlo Ciliberto
MeLa infers shared labels across few-shot tasks whose class identifiers are task-local, enabling supervised pre-training without a supplied global label map. It learns features, clusters local classes while enforcing within-task label constraints, and pre-trains on inferred labels. On H-Aircraft, it reaches 84.8 ± 0.3% one-shot accuracy; its five-shot accuracy is 92.9 ± 0.2%, near the approximate oracle’s 93.1 ± 0.2%. These results do not guarantee correct labels for arbitrary concepts.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.