Reproduce “Counterfactual Reasoning for Out-of-distribution Multimodal Sentiment Analysis” 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
Teng Sun, Wenjie Wang, Liqiang Jing, Yiran Cui, Xuemeng Song, Liqiang Nie
CLUE estimates text-only shortcut effects and subtracts them from predictions using text, audio, and video, treating the remainder as its intended indirect effect. On MOSEI’s OOD negative/positive task, MAG-BERT+CLUE raises Acc-2 from 76.41% to 80.51%. The test sets construct word–sentiment shifts, and the comparison does not isolate causal intervention.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.