Reproduce “BlendX: Complex Multi-Intent Detection with Blended Patterns” 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
Yejin Yoon, Jungyeon Lee, Kangsan Kim, Chanhee Park, Taeuk Kim
BlendX tests multi-intent detection beyond predictable joins with varied connectors and blended phrasing that can leave information implicit or refer back. It extends four datasets using rule-based and ChatGPT-assisted construction. On SNIPS, TFMN accuracy falls from 95.68 ± 0.57 on MixX to 52.51 ± 1.86 on BlendX, but reaches 94.93 ± 0.85 when trained and tested there. This shows distribution-shift sensitivity, not real-world prevalence or deployment benefit.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.