Abstract
Latent world models predict future states for goal-directed planning using action chunks spanning multiple primitive steps. Existing methods typically use fixed-length chunks and either omit goal-conditioned action generation or limit their supervision to short goal spans. We introduce FlexiWorld, a JEPA-based world model that combines mixed-span goal supervision with variable-length action chunks to improve long-horizon control. During training, we sample varying goal spans and randomly partition the actions into variable-length chunks. We jointly train the world model with a causal action encoder that embeds variable-length chunks and an autoregressive actor that generates primitive actions sequentially. Student Forcing reduces exposure bias by training on generated action prefixes. For planning, Actor-Residual Cross-Entropy Method (ARCEM) combines action-residual search with within-chunk autoregressive feedback and chunk-boundary latent prediction. Across four benchmarks and goal distances, FlexiWorld with ARCEM achieves 89.29% mean success, compared with 83.98% for the strongest baseline. PushT ablations show improved direct control from mixed-span supervision, variable-length chunks, and Student Forcing. Without retraining, FlexiWorld supports different planning chunk lengths: longer chunks accelerate ARCEM by approximately $1.3\times$ on average while maintaining comparable average success.
Keywords
Subject
Publication details
- Journal
- Not available
- Open access
- Green open access
Cite this article
APA 7
Ren, S., Gu, Q., Ni, Z., Sun, J., Wang, J., Scieur, D., & Liu, Y. (2026). FlexiWorld: Learning and Planning via Flexible Action Chunks Across Multiple Time Scales. https://omanscience.com/en/articles/flexiworld-learning-and-planning-via-flexible-action-chunks-across-multiple-time-scales
MLA 9
Ren, Shidu, et al. "FlexiWorld: Learning and Planning via Flexible Action Chunks Across Multiple Time Scales." https://omanscience.com/en/articles/flexiworld-learning-and-planning-via-flexible-action-chunks-across-multiple-time-scales.
Chicago (author–date)
Ren, Shidu, Qilin Gu, Zhenghao Ni, Junhan Sun, Jiaqi Wang, Damien Scieur, and Yunze Liu. 2026. "FlexiWorld: Learning and Planning via Flexible Action Chunks Across Multiple Time Scales." https://omanscience.com/en/articles/flexiworld-learning-and-planning-via-flexible-action-chunks-across-multiple-time-scales.
Harvard
Ren, S., Gu, Q., Ni, Z., Sun, J., Wang, J., Scieur, D. and Liu, Y. (2026) 'FlexiWorld: Learning and Planning via Flexible Action Chunks Across Multiple Time Scales', Available at: https://omanscience.com/en/articles/flexiworld-learning-and-planning-via-flexible-action-chunks-across-multiple-time-scales.
Vancouver
Ren S, Gu Q, Ni Z, Sun J, Wang J, Scieur D, et al. FlexiWorld: Learning and Planning via Flexible Action Chunks Across Multiple Time Scales. https://omanscience.com/en/articles/flexiworld-learning-and-planning-via-flexible-action-chunks-across-multiple-time-scales
IEEE
S. Ren, Q. Gu, Z. Ni, J. Sun, J. Wang, D. Scieur, and Y. Liu, "FlexiWorld: Learning and Planning via Flexible Action Chunks Across Multiple Time Scales," https://omanscience.com/en/articles/flexiworld-learning-and-planning-via-flexible-action-chunks-across-multiple-time-scales.