Reproduce “LEV: Thinking Fast and Deep with Looped Decision 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
Dingjie Song, Zhiling Yan, Yuxuan Zhang, Lichao Sun
The authors present LEV, a decision model that assigns probabilities to options and runs at several depths on one checkpoint. It repeats shared layers, scoring options after each step without generating an explanation. On 231 JevBench questions, LEV-2.6B accuracy rises from 63.2% at depth one to 76.2% at depth four, with the same weights and gains concentrated on harder questions.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.