Reproduce “Leap-LSTM: Enhancing Long Short-Term Memory for Text Categorization” 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
Ting Huang, Gehui Shen, Zhi-Hong Deng
Leap-LSTM classifies whole documents by deciding at each word whether to update its LSTM state or copy the previous state. It bases each decision on preceding, current, and following-text information. On AGNews, at a desired skip rate of 0.6, it skips 57.08% of words, reaches 93.64% accuracy, and reports 1.5× test-time speedup; results vary by task and skip rate.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.