Reproduce “Interpretable Text Embeddings Based on Hierarchical Questions” 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
Kohei Oda, Po-Min Chuang, Kiyoaki Shirai, Natthawut Kertkeidkachorn
The authors organize QA-based text embeddings into hierarchies of broad and specific yes/no questions, then align lower-level questions with broader parent questions using an NLI model. Across semantic textual similarity, clustering, and retrieval benchmarks, alignment improves the reported performance–cognitive-load trade-off, where cognitive load is a question-count proxy. The results do not show that people interpret embeddings more easily.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.