Reproduce “Auditing Quality Filters for Long-Tail Human Data Curation” 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
Rishav Agarwal, Nirshal Chandra Sekar, Anirudh Vemula
The authors examine how pose-quality filtering and person detection may reduce representation of low-pose construction workers in curated data. They audit filter survival using four alternative proxies, then test detection in simulations with YOLO11s and height-matched figures. Shrinking standing figures leaves a 17.7-percentage-point recall gap, but the simulated result neither establishes a real-site effect nor isolates posture as its cause.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.