Reproduce “EXPLORE-Bench: Egocentric Scene Prediction with Long-Horizon Reasoning” 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
Chengjun Yu, Xuhan Zhu, Chaoqun Du, Pengfei Yu, Wei Zhai, Yang Cao, Zheng-Jun Zha
EXPLORE-Bench evaluates whether models can predict a final scene from a first-person image and long sequence of atomic actions. It contains 1,157 instances with final-scene annotations for objects, attributes, and relations. On a 100-instance subset, humans score 59.08 versus Qwen3-VL-8B-Thinking’s 51.69 on a weighted unified score; multi-turn inference raises Long-subset score by 3.41 points, with increased inference time under selected settings.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.