Reproduce “Path-Level Structural and Semantic Alignment for Hierarchical Multi-Label Text 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

Xindi Wang, Aifen Xiong, Baiyang Wang, Xuri Ge, Robert Mercer, Frank Rudzicz, Yifan Peng, Zhumin Chen and 1 more

The authors train hierarchical multi-label text classifiers with complete root-to-label paths and document-conditioned descriptions. These objectives supplement ordinary label-wise training and are removed at inference. Across three runs on NYT, the full model reports Micro-F1 82.52 ± 0.22, versus 81.79 ± 0.18 for BCE-only and 82.29 ± 0.12 without descriptions; this comparison is not significance-tested.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.