Modelsslug: lev

Lev

Adaptive decision model that deepens reasoning via recurrent loops instead of generated tokens.

Curated Intelligence

1 items · Page 1/1
新智元 (微信公众号)T2·Models
62 pts
Lev Loops Instead of Tokens to Scale Reasoning Depth

Researchers from Lehigh University, UBC and the Vector Institute propose Lev, a decision model that scales reasoning depth through recurrent loops rather than generated reasoning tokens, keeping simple requests at tens of milliseconds while giving hard ones more compute. It reports up to 13% accuracy gains on decision tasks, with code, models and datasets released openly.