Reproduce “TSEmbed: Unlocking Task Scaling in Universal Multimodal Embeddings” 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
Yebo Wu, Feng Liu, Ziwei Xie, Zhiyuan Liu, Changwang Zhang, Jun Wang, Li Li
TSEmbed adapts a multimodal embedding model for image and text matching across tasks by routing inputs through specialist low-rank updates. It first stabilizes routing with contrastive training, then uses routing overlap to weight hard negatives. On MMEB, the 7B model scores 74.7 overall versus 72.0 for B3; this benchmark result does not establish routing overlap as general semantic similarity.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.