Latent Action as Intention Enables Efficient Future Imagination for World Action Models
Summary
LAWA is a world action model that uses latent actions to enable efficient future imagination for robot control, achieving state-of-the-art performance with reduced inference latency. It improves over baselines in generalization and efficiency without generating future observations.
View Cached Full Text
Cached at: 08/26/26, 11:11 AM
Paper page - Latent Action as Intention Enables Efficient Future Imagination for World Action Models
Source: https://huggingface.co/papers/2608.24882 Authors:
,
,
,
,
,
,
,
,
,
,
,
Abstract
LAWA improves robot control by using compact latent actions to retain efficient future imagination without generating observations, achieving strong performance with lower latency.
World action models(WAMs) improve robot control by modeling how observations evolve, but generating future observations at test time incurs substantial latency. Fast-WAM removes this process for efficiency; however, our matched implementations show lower generalization for Fast-WAM than for future-aware alternatives, especially with scarce robot demonstrations and in out-of-distribution scenarios. To bridge this gap, we introduce **LAWA**, a WAM architecture that uses compactlatent actionsas an operational representation of future intentions, enabling efficient test-time future imagination without generating future observations. Specifically, adiscrete tokenizerenhanced byaction-free pre-trainingproduces manipulation-centriccodebooktargets. LAWA jointly denoises a continuouslatent stateanchored to these targets with executableaction chunkswhile omitting thefuture-video branchat inference. On RoboCasa, LAWA achieves state-of-the-art average success rates of 65.6% and 80.8% in the few-shot and full data settings, improving over the matched Fast-WAM baseline by 9.6 and 4.5 points, respectively. It also preserves the performance level of the matched Joint-WAM variant while requiring 42.9% lowerinference latency. LAWA also demonstrates competitive zero-shot robustness on LIBERO-Plus and superior performance on real-world tasks. These results show that future imagination need not be discarded: retaining it with compactlatent actionsyields an effective trade-off among performance, generalization, and latency. Code and models will be released.
View arXiv pageView PDFProject pageAdd to collection
Get this paper in your agent:
hf papers read 2608\.24882
Don’t have the latest CLI?curl \-LsSf https://hf\.co/cli/install\.sh \| bash
Models citing this paper0
No model linking this paper
Cite arxiv.org/abs/2608.24882 in a model README.md to link it from this page.
Datasets citing this paper0
No dataset linking this paper
Cite arxiv.org/abs/2608.24882 in a dataset README.md to link it from this page.
Spaces citing this paper0
No Space linking this paper
Cite arxiv.org/abs/2608.24882 in a Space README.md to link it from this page.
Collections including this paper0
No Collection including this paper
Add this paper to acollectionto link it from this page.
Similar Articles
LaWAM: Latent World Action Models for Efficient Dynamics-Aware Robot Policies
LaWAM enables efficient robot control by predicting compact latent visual subgoals instead of expensive video generation, achieving state-of-the-art success rates with up to 24x lower latency than pixel-space world action models.
Foresight Without Seeing: Latent Futures for World Action Models
Presents ForeWAM, a direct-policy World Action Model that provides predictive context to action generation via hidden future-slot KV states, avoiding explicit future video decoding while achieving high success rates on LIBERO benchmarks.
@aimalysheva: latent actions are having a moment, especially in robotics: instead of predicting a robot's actual joint commands or ga…
Latent actions are gaining traction in robotics as a way to learn from unlabeled video without action labels. Recent papers from DeepMind and FAIR demonstrate progress from controlled game environments to in-the-wild internet video, promising scalable training for imitation learning.
World-Language-Action Model for Unified World Modeling, Language Reasoning, and Action Synthesis
This paper introduces World-Language-Action (WLA) models, embodied foundation models that jointly predict textual subtasks, subgoal images, and robot actions from text, images, and robot states, achieving state-of-the-art multi-task and long-horizon learning in simulated and real-world environments.
Predicting Consequences and Reinforcing Navigation Policies with Latent World Models
This paper proposes a Latent World Model (LWM) for robot navigation that predicts action-conditioned latent feature compatibility, enabling policy learning from unlabeled video data and reinforcement learning without additional environment interaction, outperforming existing methods.