DISK: Dynamic Inference SKipping for World Models

Anugunj Naman, Gaibo Zhang, Ayushman Singh, Yaguang Zhang · 2026

Read DISK: Dynamic Inference SKipping for World Models by Anugunj Naman, Gaibo Zhang, Ayushman Singh, Yaguang Zhang (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-2602-00440.txt.

Description

Anugunj Naman and team present DISK, a training-free inference method for autoregressive world models. DISK links two diffusion transformers for video and ego paths via dual controllers with skip decisions, keeping motion consistency without retraining. It extends latent skip testing into the autoregressive chain-of-forward regime and passes controller statistics through rollouts for stability. In driving rollouts on 1500 samples, DISK hits 2x speedup on trajectory diffusion and 1.6x speedup on video diffusion. Planning error, visual quality, and NAVSIM scores hold, showing practical long horizon prediction at reduced cost.

Catalog record

CREATORAnugunj Naman, Gaibo Zhang, Ayushman Singh, Yaguang Zhang
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE33397
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2602.00440

Browse the full library · About the Tower of Babel