Skip to content
Preprint

AWM-VLA: AlignedWorld Modeling for Efficient and Explainable Vision-Language-Action Policies

Aug 2026 · 0 citations · 42 references
Computer Science

TL;DR

AWM-VLA is presented, a unified framework that embeds aligned world modeling directly inside a diffusion-transformer policy and is compatible with any diffusion or flow-matching policy, making aligned world modeling an inexpensive, broadly applicable component of generalist manipulation.

Abstract

Vision-language-action (VLA) models have become a powerful paradigm for generalist robotic manipulation, yet they are often reactive: the policy maps the current observation directly to an action chunk without reasoning about the long-term consequences of its decisions. Prior attempts to endow policies with world models either reconstruct future frames in pixel space---expensive and dominated by task-irrelevant detail---or decouple the world model from the policy, weakening control. We present AWM-VLA, a unified framework that embeds aligned world modeling directly inside a diffusion-transformer policy. Following the Future Latent REpresentation Alignment (FLARE) principle, we add learnable future tokens whose intermediate activations are aligned with vision-language embeddings of future observations, enabling the policy to anticipate long-term consequences while generating actions. We extend this paradigm in two ways. First, we introduce an object-centric decoupled alignment objective that predicts future object-level semantics alongside the global future embedding, improving both interpretability and multi-instruction generalization. Second, we balance the global and object-centric alignment terms against the action flow-matching loss through a principled weighting, yielding a controllable accuracy--interpretability trade-off. On RoboCasa and humanoid tabletop manipulation benchmarks, AWM-VLA outperforms prior VLA and world-model baselines by up to 21% in success rate, improves generalization to novel objects and instructions, and produces object-centric rationales that are preferred by human raters in 83 of cases. Our approach adds only a few learnable tokens to the policy and is compatible with any diffusion or flow-matching policy, making aligned world modeling an inexpensive, broadly applicable component of generalist manipulation.

View source

Similar papers

Preprint Aug 2026

World Tokens: Enhancing Embodied Policies with Training-Time World Modeling

World Tokens is an embodied policy architecture built around a World Adapter that bridges visual-language understanding, world-dynamics modeling, and action generation and is highly competitive on LIBERO, attains the best reported averages on SIMPLER, and substantially improves real-world R1 Pro success over a matched...

Qu Tang, Benhui Zhuang, Bo Yuan et al. · 2 citations · ⚡1
Sep 2026

SLIP-VLA: Single-Step Latent Imagination for Policy Learning in Vision-Language-Action Models

SLIP-VLA, a policy learning framework that equips VLA models with a Single-Step Latent Imagination for future-aware action prediction, achieves state-of-the-art performance across diverse simulation benchmarks and real-world manipulation tasks, while its single-step latent imagination takes only 12 ms.

Tian-Fu Li, Hao-Xuan Xu, Wen-Bo Chen et al. · 0 citations
#small language model Conference Open access Jul 2026

Reasoning Without Inference Cost: Latent Semantic Scaffolding for Robot VLA Policies

This work introduces Latent Semantic Scaffolding (LSS), an auxiliary loss applied during human-demonstration pretraining that aligns a VLA's action-token representations to text embeddings of physical-reasoning rationales through a small projection head.

A. Li, Zhuo Li, Zhe-Lin Yang et al. · 3 citations
Preprint Oct 2026

UniWAM: Unified World-Action Model

UniWAM is introduced, a unified architecture that integrates a physical reasoner, a world generator, and an action predictor to jointly learn semantic understanding of the physical world, visual generation, and action prediction and uncovers a log-linear scaling law of unified human-robot co-training.

Wen-Xuan Song, Jia-Yi Chen, Jing-Bo Wang et al. · 6 citations · ⚡2
Preprint Aug 2026

Temporal Forcing: 4D Representation Alignment for Vision-Language-Action Models

This work introduces a history pathway that enables a vanilla VLA model to summarize observation history into temporally aware latent representations, which captures the evolving 3D world through temporally consistent geometric representations, enabling a deeper understanding of dynamic environments.

Xing-Yu Ding, Yu-Zhong Zhao, Chun-Ming Zhao et al. · 1 citation

We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.