Robotics · 2026-10-14
WAVE-0: Bridging VLA Policies and WAMs with Lightweight Latent Future Prediction
Haomin Wang, Anqiong Ji, Ke Wang, Zhongshan Liu, Ru Ying, Ran Zheng, Zhipeng Zhang
AI-assisted summary based on the paper. Consult the linked source for methods and limitations. Benchmark numbers appear after editor verification.
TL;DR
WAVE-0 (World-model-Augmented VLA Execution) bridges vision-language-action (VLA) policies and world action models (WAMs) with a lightweight latent world model. Using a 257M action-conditioned latent model trained directly on ground-truth robot actions with a counterfactual ranking objective, it predicts future representations that condition the action expert. It reports high success on RoboTwin 2.0 and LIBERO, plus gains on real-world tasks.
Why it matters
The authors address that VLA policies do not explicitly model action consequences, while many WAMs rely on costly video generation; WAVE-0 replaces pixel/video prediction with action-grounded latent future prediction. By coupling latent planning (from a VLM) to future-conditioned action generation via an explicit latent world model supervisor, the method aims to improve success and generalization without large-scale video pretraining.
Method
- Train a lightweight latent world model in Stage 1 to predict future visual representations in a fixed embedding space from current observation, proprioceptive state, and ground-truth actions, using future-representation loss, latent delta loss, and a counterfactual ranking objective.
- In Stage 2, integrate the frozen latent world model into pretrained π0.5: train the VLM planner to output latent action plans aligned with the learned action representation space, then use a resampler to convert predicted future representations into conditioning tokens for the action expert.
- During training and inference, action generation is guided by predicted consequences (future-conditioned tokens), while the expert retains the original VLM context; the approach avoids pixel reconstruction and large-scale video pretraining.
Abstract (from arXiv)
Vision-language-action (VLA) policies excel at generalist robotic manipulation but generate actions without explicitly modeling their environmental consequences. World action models (WAMs) incorporate future prediction, yet most rely on costly video generation, while latent alternatives either use predicted futures only as auxiliary targets or infer latent actions from suboptimal visual transitions through costly large-scale video pretraining. We introduce WAVE-0 (World-model-Augmented VLA Execution), a framework that bridges VLAs and WAMs through a lightweight latent world model. By connecting the semantic understanding of pretrained VLAs with explicit future prediction, WAVE-0 allows anticipated action consequences to directly guide action generation. Built on a pretrained VLA, its 257M world model translates VLM-generated latent action plans into predicted future representations that condition the action expert, unifying latent planning, future prediction, and action generation. The world model is trained directly on ground-truth robot actions with a counterfactual ranking objective, requiring neither pixel reconstruction nor large-scale video pretraining. WAVE-0 achieves success rates of 91.47% on RoboTwin 2.0 and 98.7% on LIBERO, along with 88.0%, 61.4%, and 75.0% on LIBERO-Plus, LIBERO-Pro, and LIBERO-Para, respectively. On real-world tasks, it improves the success rate of $\pi_{0.5}$ from 38.3% to 73.3%. These results demonstrate strong manipulation performance and generalization across diverse distribution shifts.
Related papers
- Forward Dynamics Models
- PLaW-VLA: Predictive Latent World Modeling for Vision-Language-Action Policies
- TS-WAM: Temporal Latent Actions and World-Action Coupling for Vision-Language Robot Control
- Rewiring Semantics, Dynamics, and Control: A Simple yet Effective Action-Centric Tri-Stream Transformer
- Vision-Language-Delta-Action: Learning Compact Deltas as Actionable Transition Interfaces for Robotic Manipulation
Publication record
Bibliographic evidence describes where and when a record appeared. It does not validate scientific claims.
First public release: 2026-10-13 · Indexed by Pub-AI: 2026-10-14 UTC
WAVE-0: Bridging VLA Policies and WAMs with Lightweight Latent Future Prediction
Preprint posted: 2026-10-13 · Original record ↗
- arxiv
- 2610.18568
Explicit source versions (1)
- v1 · 2026-10-13
- Preprint available
Source date: 2026-10-13 · First detected: 2026-10-25 UTC · Historical baseline
Source evidence (1)
- Official evidence ↗ · atom:published (official metadata)
Abstract · arxiv
Vision-language-action (VLA) policies excel at generalist robotic manipulation but generate actions without explicitly modeling their environmental consequences. World action models (WAMs) incorporate future prediction, yet most rely on costly video generation, while latent alternatives either use predicted futures only as auxiliary targets or infer latent actions from suboptimal visual transitions through costly large-scale video pretraining. We introduce WAVE-0 (World-model-Augmented VLA Execution), a framework that bridges VLAs and WAMs through a lightweight latent world model. By connecting the semantic understanding of pretrained VLAs with explicit future prediction, WAVE-0 allows anticipated action consequences to directly guide action generation. Built on a pretrained VLA, its 257M world model translates VLM-generated latent action plans into predicted future representations that condition the action expert, unifying latent planning, future prediction, and action generation. The world model is trained directly on ground-truth robot actions with a counterfactual ranking objective, requiring neither pixel reconstruction nor large-scale video pretraining. WAVE-0 achieves success rates of 91.47% on RoboTwin 2.0 and 98.7% on LIBERO, along with 88.0%, 61.4%, and 75.0% on LIBERO-Plus, LIBERO-Pro, and LIBERO-Para, respectively. On real-world tasks, it improves the success rate of $π_{0.5}$ from 38.3% to 73.3%. These results demonstrate strong manipulation performance and generalization across diverse distribution shifts.