Reproduce “JABER: Sequential Quality Control for Human-AI Data Annotation under Shared Bias” 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
Fengxu Yang, Minghui Liwang, Yang Lu, Bo Han, Xianbin Wang, Seyyedali Hosseinalipour
JABER manages annotation quality control by tracking whether a held label is correct and whether shared bias may explain agreement. It compares accepting, inspecting, relabeling, arbitrating, or leaving a task unresolved by their costs and future consequences. In synthetic tests, it reduced expected loss; on 40 FEVER claims, it lowered loss versus two baselines, but matched a fixed sequence.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.