Robotics · arXiv:2606.15768 v2 · submitted 14 Jun 2026
LaWAM: Latent World Action Models for Efficient Dynamics-Aware Robot Policies
The paper says
“LaWAM achieves state-of-the-art or competitive success rates (SRs) across LIBERO (98.6% SR), RoboTwin (91.22% SR), and real-world manipulation tasks while retaining low-latency inference.”
from the abstract · picked out as the main claim by a classifier
Reviews
Reviews by people
No person has reviewed this paper on Pith yet. A scientist listed on Pith can review it for a fee, and their review is published here whatever it concludes.
The paper
Abstract
Vision-Language-Action models (VLAs) leverage large-scale vision-language pretraining for semantic robot control, but often lack explicit foresight into how robot actions change the scene. World-Action Models (WAMs) address this limitation by conditioning policies on predicted futures, yet existing approaches typically rely on computationally expensive video generation with substantial pixel-level redundancy. We present LaWAM, a Latent World Action Model that exposes predictive dynamics to robot policies through compact latent visual subgoals instead of reconstructed future video. At the core of LaWAM is a latent-action-conditioned Latent World Model (LaWM). We obtain LaWM by training a latent action model in the latent space of a pretrained vision foundation model and repurposing its forward decoder to predict future observation features for scene evolution. LaWAM then conditions action generation on these predicted latent visual subgoals to enable dynamics-aware robot control. LaWAM achieves state-of-the-art or competitive success rates (SRs) across LIBERO (98.6% SR), RoboTwin (91.22% SR), and real-world manipulation tasks while retaining low-latency inference. LaWAM runs in 187 ms per action-chunk prediction and achieves up to 24x lower wall-clock latency than pixel-space WAMs.
The paper's own sentences, arranged. Each label is drawn from a closed vocabulary; the number beside it is the probability the classifier gave that label. No sentence in this record was written by a model.
abstract compiled · 2026-10-27 · no language-model review
How the main claim is stated: asserted 1.00
What the abstract itself states, as probabilities: claims first 0.07 · states limitation 0.08 · human participants 0.06 · controlled comparison 0.57 · randomized assignment 0.01 · states quantitative result 0.99 · compares with prior results 0.94 · mentions code or data availability 0.03
The abstract, sentence by sentence
This table labels the abstract only (arXiv listing abstract v2). The number is the probability the classifier gave the label.
| Sentence | Role | The authors' own work | States the main conclusion | Strength |
|---|---|---|---|---|
| Vision-Language-Action models (VLAs) leverage large-scale vision-language pretraining for semantic robot control, but often lack explicit foresight into how robot actions change the scene. | background 0.82 | 0.05 | 0.03 | asserted 0.98 |
| World-Action Models (WAMs) address this limitation by conditioning policies on predicted futures, yet existing approaches typically rely on computationally expensive video generation with substantial pixel-level redundancy. | background 0.67 | 0.05 | 0.03 | asserted 0.98 |
| We present LaWAM, a Latent World Action Model that exposes predictive dynamics to robot policies through compact latent visual subgoals instead of reconstructed future video. | contribution 1.00 | 0.89 | 0.39 | asserted 1.00 |
| At the core of LaWAM is a latent-action-conditioned Latent World Model (LaWM). | method 0.74 | 0.60 | 0.06 | asserted 1.00 |
| We obtain LaWM by training a latent action model in the latent space of a pretrained vision foundation model and repurposing its forward decoder to predict future observation features for scene evolution. | method 0.99 | 0.93 | 0.06 | asserted 1.00 |
| LaWAM then conditions action generation on these predicted latent visual subgoals to enable dynamics-aware robot control. | method 0.96 | 0.79 | 0.15 | asserted 0.99 |
| LaWAM achieves state-of-the-art or competitive success rates (SRs) across LIBERO (98.6% SR), RoboTwin (91.22% SR), and real-world manipulation tasks while retaining low-latency inference. | result 1.00 | 0.95 | 0.45 | asserted 1.00 |
| LaWAM runs in 187 ms per action-chunk prediction and achieves up to 24x lower wall-clock latency than pixel-space WAMs. | result 1.00 | 0.95 | 0.16 | asserted 1.00 |
What kind of paper the classifier reports this is: method or tool 1.00 · evidence from observation or experiment 0.96
What was examined, and what was not
Sources read
- arXiv listing abstract v2
Not examined
- full text
- figures and tables
- references and citing sentences
- relations to other papers
Levels run
- P abstract
Not run
- S sections
- T body sentences
- C citations
- R cross-paper relations
How the labels were made
Compiled by classifier-1.13.0 via TypeSafe API, question bank abstract-p1-20260923, 2026-10-27: 1 request, 6,358 input tokens.
Every quoted sentence on this record is the authors'. Every label is from Pith's closed vocabulary, with the probability shown beside it. This record as JSON
Cited by 13 Pith papers
Reviewed papers on Pith that cite this one. The full citation record
- Beyond Prediction Gain: LASR as a Metric for Action-Grounded Latents in World Models cs.RO · Nov 2026
- Toward the Era of Agentic Embodied Intelligence cs.RO · Nov 2026
- Gnomon4D: Learning Where and When to Focus for Geometric Motion-Grounded World-Action Modeling cs.CV · Nov 2026
- AdaPolicy: Action-Driven Anticipatory Robot Policies via Next-Chunk Prediction cs.RO · Oct 2026
- DiffWAM: Learning World Action Model from Consecutive Visual Difference cs.RO · Nov 2026
- Be My Eye: Learning to See Beyond a Robot's Own Perspective cs.RO · Nov 2026
- ThinkWorld: World Imagination Reinforces Chain-of-Dynamics Reasoning cs.RO · Oct 2026
- SO-101 Bench: Accessible Spatial Manipulation cs.CV · Oct 2026
- Pretraining Knows What, Simulation Teaches How: Compositional Generalization in a Sim-Trained Robot Agent cs.RO · Oct 2026
- Asynchronous Robot Policies cs.RO · Nov 2026
- Continuous-Action World Models: Learning Cross-Embodiment Spatiotemporal Dynamics Priors cs.RO · Oct 2026
- Trace-WAM: Learning to Act via 3D Motion Prediction for World Action Models cs.RO · Oct 2026
- From Egocentric Human Video to Robot Manipulation in the Era of Embodied Intelligence cs.RO · Nov 2026
How the literature uses it
20 arXiv papers cite this in 50 sentences
Papers citing it · 30 sentences without a confident label
Builds on it 1 paper
-
LaWAM showed that a latent visual subgoal can preserve the control benefit of future prediction without reconstructing pixels [ref] . builds on this result by exposing the policy/world boundary in the implementation and evaluating a compact policy configuration.
AcrossWAM1.0:A Modular Latent World-Action Stack for Compact Robot Policies · Related Work · 0.91
Uses its method or tool 1 paper
-
To further encourage these implicit future states to focus on scene transitions induced by robot interaction, we introduce dynamics registers supervised by a frozen LaWM latent-action teacher [ref] .
Foresight Without Seeing: Latent Futures for World Action Models · Introduction · 0.91
Compares with it 3 papers
-
For additional context, the four-task LaWAM + LEON model also remains competitive with the corresponding published four-task slice of LaWAM [ref] trained on the full 50-task mixture, achieving 84.13% versus 83.50% in combined performance.
Making Latent Evolution Explicit: Operator-Structured Transitions for World Action Models · Policy Performance Across Prediction–Policy Couplings · 0.96
-
The main comparison includes Active Neural Time Fields [ref] , Generic world models (the capacity matched generative action-consistent world model) [ref] with the MPC budget, and the complete ELWM+PC-NTF system.
Energy-Structured Latent World Models with Neural Time Fields for Physically Constistent Open-World Motion Planning · Experimental Setup · 1.00
-
These include VLAs such as $π_0$ [ref] , $π_0.5$ [ref] ; video-generative WAMs such as Fast-WAM [ref] , Motus [ref] , and LingBot-VA [ref] ; and methods exploring alternative future representations, such as LaWAM [ref] , Mask World Model [ref] , MaskWAM [ref] .
ST-WAM: Semantic-Temporal World Action Model for Robust Manipulation under Visual Distribution Shifts · Experimental Setup · 0.98
Cites it as background 10 papers
-
A parallel line of work reduces the deployment cost: FastWAM [ref] questions whether test-time future imagination is needed, AHA-WAM [ref] amortizes world planning across multiple action updates through asynchronous execution, and LaWAM [ref] exposes predictive dynamics through compact latent visual subgoals instead of future video.
Rethink Before You Execute: Adaptive Execution for World Action Models · Related Work · 0.97
-
World Action Models (WAMs) incorporate such future prediction into robot policy learning and control [ref] .
Making Latent Evolution Explicit: Operator-Structured Transitions for World Action Models · Introduction · 0.88
-
LDA-1B first scales this into a semantic-latent world-action model that jointly denoises future DINO states and action chunks [ref] , and later work extends this semantic-latent world-action modeling [ref] .
Robust-WAM: Bridging Generative Pretraining and Semantic Foresight in World-Action Models · Related Work · 0.84
Challenges and discussion
Know something about this paper? Leave a comment.
Come back by email
A short list of the week's newly reviewed papers. Nothing is sent until you confirm from your inbox, and every email has a link that stops it. No account.
No comments yet.
Sign in to comment