Reproduce “Multimodal Fusion with LLMs for Engagement Prediction in Natural Conversation” 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

Cheng Charles Ma, Kevin Hyekang Joo, Alexandria K. Vail, Sunreeta Bhattacharya, Álvaro Fernández García, Kailana Baker-Matsuoka, Sheryl Mathew, Lori L. Holt and 1 more

The authors test whether GPT-4 can predict participants’ self-reported conversational engagement by turning wearable recordings into transcripts annotated with gaze and facial-expression cues, then prompting the model with personal context. In 34 participants across 19 dyads, its full-input version had mean RMSE 1.364, close to but above the SVM’s 1.327; this preliminary result concerns self-reports in a small sample.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.