Driving · 2026-10-12
Forward Dynamics Models
Zipei Ma, Junzhe Jiang, Jiahui Zhang, Juntong Zhang, Chun Gu, Bozhou Zhang, Jiankang Deng, Xiatian Zhu, Li Zhang
AI-assisted summary based on the paper. Consult the linked source for methods and limitations. Benchmark numbers appear after editor verification.
TL;DR
Forward Dynamics Models (FDM) couple action learning with future world-state prediction using an action-to-future paradigm: future prediction reads action features, while action prediction cannot access future information. End-to-end training uses ground-truth future states to supervise the policy through action features, so prediction errors cannot propagate into actions. At inference, FDM skips future generation, and is instantiated with both VLM and VGM backbones.
Why it matters
FDM targets two issues of future-conditioned action models: extra inference cost from future generation and error propagation from visual prediction into actions. By reversing the conditioning direction (action-to-future) and backpropagating future loss through action features, the authors report improved task success while enabling action-only inference without future generation.
Method
- Action-to-future asymmetric conditioning: future prediction reads action features; action prediction cannot access future information (training couples action imitation with future world-state prediction).
- End-to-end training backpropagates future-prediction gradients through action features to supervise the policy; action errors cannot propagate from predicted futures into actions.
- Action-only inference: the future-generation branch is skipped, avoiding computational overhead from future generation while retaining context processing.
Abstract (from arXiv)
Vision-language-action (VLA) models and world action models (WAMs) have advanced robotic policy learning by building on pretrained vision-language models (VLMs) and video generative models (VGMs), respectively. Yet VLAs learn without explicit supervision from future world states, while WAMs that condition actions on predicted futures incur extra inference cost and let visual prediction errors propagate into actions. We present Forward Dynamics Models (FDM), a family of robotic action models that couples action and future world-state prediction through an action-to-future paradigm. Future prediction is conditioned on action features, never the reverse: end-to-end training lets ground-truth future states supervise the policy through these features, while prediction errors cannot propagate into actions. At inference, FDM skips future generation entirely, avoiding its computational cost. The paradigm is backbone-agnostic, and we instantiate it with both VLM and VGM backbones. Experiments in simulation and on real robots demonstrate the effectiveness of FDM across both backbone families, achieving a state-of-the-art 99.4% average success rate on LIBERO and absolute gains of up to 13% over VLA and WAM methods on real-world tasks.
Related papers
- WAVE-0: Bridging VLA Policies and WAMs with Lightweight Latent Future Prediction
- De-Diffusing WAMs: Action-Sufficient Direct Prediction for Robot Control
- MiniWAM: Learning Compact Future Targets for Efficient World-Action Modeling
- FutureWorlds: Learning Robotic World Models from Alternative Futures
- What Should Robot Policies Learn to Predict? Rethinking Supervision Targets in Vision-Language-Action Learning
Publication record
Bibliographic evidence describes where and when a record appeared. It does not validate scientific claims.
First public release: 2026-10-09 · Indexed by Pub-AI: 2026-10-12 UTC
Forward Dynamics Models
Preprint posted: 2026-10-09 · Original record ↗
- arxiv
- 2610.13117
Explicit source versions (1)
- v1 · 2026-10-09
- Preprint available
Source date: 2026-10-09 · First detected: 2026-10-25 UTC · Historical baseline
Source evidence (1)
- Official evidence ↗ · atom:published (official metadata)
Abstract · arxiv
Vision-language-action (VLA) models and world action models (WAMs) have advanced robotic policy learning by building on pretrained vision-language models (VLMs) and video generative models (VGMs), respectively. Yet VLAs learn without explicit supervision from future world states, while WAMs that condition actions on predicted futures incur extra inference cost and let visual prediction errors propagate into actions. We present Forward Dynamics Models (FDM), a family of robotic action models that couples action and future world-state prediction through an action-to-future paradigm. Future prediction is conditioned on action features, never the reverse: end-to-end training lets ground-truth future states supervise the policy through these features, while prediction errors cannot propagate into actions. At inference, FDM skips future generation entirely, avoiding its computational cost. The paradigm is backbone-agnostic, and we instantiate it with both VLM and VGM backbones. Experiments in simulation and on real robots demonstrate the effectiveness of FDM across both backbone families, achieving a state-of-the-art 99.4% average success rate on LIBERO and absolute gains of up to 13% over VLA and WAM methods on real-world tasks.