Reproduce “Autoregressive Score Generation for Multi-trait Essay Scoring” 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
Heejin Do, Yunsu Kim, Gary Geunbae Lee
ArTS uses a T5 encoder–decoder model to generate a sequence of named trait scores for an essay, allowing later scores to condition on earlier ones. It was evaluated with five-fold cross-validation on ASAP/ASAP++, using quadratic weighted kappa. Across traits, ArTS averaged 0.695 QWK versus 0.638 for MTL-BiLSTM. The gain was not uniform: Overall and Voice scored lower than the baseline.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.