Reproduce “Disambiguating Reference in Visually Grounded Dialogues through Joint Modeling of Textual and Multimodal Semantic Structures” 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
Shun Inadumi, Nobuhiro Ueda, Koichiro Yoshino
The authors present a model that trains textual relations among dialogue mentions alongside links between mentions and image objects, enabling textual cues to inform visual grounding. On J-CRe3, coreference-trained MRR reaches Recall@1 of 0.361 for 120 pronoun instances, versus 0.277 for its MRR-only baseline. This evidence is limited to Japanese dialogue grounding, not robot task success.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.