Reproduce “Intern-Decision: Efficient Multimodal Decision Making with Autoregressive Masked Language Modeling” 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
Tong Zhu, Zhe Xu, Zefeng He, Yunkai Chai, Xuyang Hu, Dehong Zheng, Yun Luo, Yafu Li and 2 more
Intern-Decision scores multiple typed choices in one multimodal model pass, returning structured labels and probabilities. It places decision placeholders in a JSON schema and reads legal-option scores from logits before each placeholder, avoiding answer-by-answer decoding. Across seven benchmark suites, Intern-Decision-4B averages 90.02%, versus Jev’s 88.74%; this is an evaluated-systems comparison, not evidence that joint scoring caused the gain.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.