Reproduce “Does System One Transfer? Calibrated Typed-Decision Models, Label Budgets, and Translation for Indonesian Regional Languages” 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
Romi Fadli, Rusdah, Wahyu Ramadani Putri Utami, Zulfa Malika Amana
The authors test Jev, a fixed-choice classifier, on paired Indonesian regional-language sentiment and emotion benchmarks. Accuracy is associated with lexical overlap with Indonesian, while human Indonesian source text can improve results. On NusaTranslation emotion, routing 30% of cases to IndoBERT in a validation-selected cascade reaches macro-F1 0.676, above IndoBERT alone. Jev answers every item; this result is task-specific.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.