Reproduce “Retrieval-augmented Encoders for Extreme 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
Yau-Shian Wang, Wei-Cheng Chang, Jyun-Yu Jiang, Jiong Zhang, Hsiang-Fu Yu, S. V. N. Vishwanathan
RAE-XMC augments a dual encoder with a joint searchable memory of labeled training examples and label descriptions, combining retrieved evidence to score labels without a trainable one-versus-all classifier. On LF-AmazonTitles-1.3M, it trained in 7.4 hours versus DEXML’s 132 hours on the same eight-A100-GPU setup. P@1 was 58.48 versus 58.43, a non-significant difference; P@5 and R@100 were significantly higher.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.