Reproduce “Evaluating Social Reasoning in Embodied Vision-Language Models” 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
Daniel Weiner, Raj Korpan
The authors evaluate Gemini Robotics-ER v2.0 on SHREC’s robot-behavior tasks, comparing transcript-only results with published GPT-4o-mini baselines. Robotics-ER reaches 0.296 detection accuracy versus 0.26 ± 0.02, but 0.382 partial-match social-attribute accuracy versus 0.41 ± 0.04. These benchmark labels concern robot behavior in human–robot interactions, not human-state detection or coordinated robot-team decisions.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.