Reproduce “ABRR: Transforming Temporal Sentence Grounding with Wasserstein-Adaptive Refinement and Precision Ranking” 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
Xiang Fang, Wanlong Fang, Junhao Dong, Daizong Liu
ABRR refines temporal sentence grounding predictions after a base model produces candidate video intervals. It smooths temporal scores, uses optimal transport to adjust boundaries, and ranks candidates by localization precision and semantic relevance without retraining. On Charades-STA with UniVTG, Recall@[email protected] rises from 29.4% to 34.2%, and mIoU from 54.3% to 58.7%, demonstrating gains for this tested model and dataset.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.