Reproduce “Reasoning-Aware Relational Representation Learning for Open-Vocabulary Scene Graph Generation” 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
Jiawei Xiao, Boya Wang, Longtian Qiu, Shan Ning, Xuming He
RaRe uses learned embeddings to connect object-pair reasoning with open-vocabulary relation retrieval. It trains on rationale–relation pairs, then rewards responses for being distinctive across object pairs before matching them to predicate descriptions. On VG150 PredCls, RaRe reaches 22.49% novel mR@50, averaging recall across predicate categories, versus 15.70% for PE-NET+RAHP. This does not extend to total recall, and gains vary by setting.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.