Describe-Then-Act: Proactive Agent Steering via Distilled Language-Action World Models

Massimiliano Pappa, Luca Romani *, Valentino Sacco *, Alessio Palma, Stéphane Lathuilière, Fabio Galasso, Xavier Alameda-Pineda, Indro Spinelli · 2026

Read Describe-Then-Act: Proactive Agent Steering via Distilled Language-Action World Models by Massimiliano Pappa, Luca Romani *, Valentino Sacco *, Alessio Palma, Stéphane Lathuilière, Fabio Galasso, Xavier Alameda-Pineda, Indro Spinelli (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-2603-23149.txt.

Description

Pappa and co-authors present DILLO, a fast steering layer for safety-critical agents. World models can foresee action outcomes, but visual simulation is slow. Their claim: a policy's latent state plus planned actions already hold enough signal. A vision-language teacher labels offline trajectories with outcome text. A language model student then learns to predict those outcomes from latent data alone. At run time the path is text-only, running many times faster. Tests on MetaWorld and LIBERO show better episode success from the steering.

Catalog record

CREATORMassimiliano Pappa, Luca Romani *, Valentino Sacco *, Alessio Palma, Stéphane Lathuilière, Fabio Galasso, Xavier Alameda-Pineda, Indro Spinelli
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE55450
LICENSEhttp://creativecommons.org/licenses/by-sa/4.0/
RIGHTSCC BY SA
SOURCEarXiv

Source: https://arxiv.org/abs/2603.23149

Browse the full library · About the Tower of Babel