Implicit Maximum Likelihood Estimation for Real-time Generative Model Predictive Control

Grayson Lee, Minh Bui, Shuzi Zhou, Yankai Li, Mo Chen, Ke Li · 2026

Read Implicit Maximum Likelihood Estimation for Real-time Generative Model Predictive Control by Grayson Lee, Minh Bui, Shuzi Zhou, Yankai Li, Mo Chen, Ke Li (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2603-13733.txt.

Description

Grayson Lee and co-authors test Implicit Maximum Likelihood Estimation, or IMLE, as a planner for real-time model predictive control. Diffusion planners capture rich, multimodal behavior, but their iterative denoising is slow. Slow plans suit closed-loop control badly, since plans must refresh fast as the scene changes. IMLE keeps strong mode coverage and runs inference two orders of magnitude faster, the team shows. On offline reinforcement learning benchmarks it matched a diffusion planner, with much faster planning. A human navigation test ran in real time.

Catalog record

CREATORGrayson Lee, Minh Bui, Shuzi Zhou, Yankai Li, Mo Chen, Ke Li
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE39140
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2603.13733

Browse the full library · About the Tower of Babel