AI Paper Atlas

Research map / Vision-language and multimodal / Robot action and navigation

ALAM: Algebraically Consistent Latent Transitions for Vision-Language-Action Models

Zuojin Tang, Haoyun Liu, Xinyuan Chang, Changjie Wu, Dongjie Huo, Yandan Yang, Bin Liu, Zhejia Cai, Feng Xiong, Mu Xu, jiachen luo, De Ma and 2 more

NeurIPS 2026 · Vision-language and multimodal › Robot action and navigation · approach: Vision-language-action foundation models

Read the paperCheck your idea against 137,217 papers

AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Proposes ALAM, a latent action model for VLA pretraining from action-free videos whose latent transitions are algebraically consistent, so the latent codes better support policy generation.

Problem

VLA models are limited by scarce action-labeled robot data, while action-free videos are abundant; reconstruction-trained latent action codes may predict future observations but lack the structure needed for policy generation.

Approach

Enforces algebraic consistency on latent action transitions learned from action-free videos, so latent codes compose structurally rather than only supporting future-observation reconstruction.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add ALAM latent transitions co-generated with actionspi0 · MetaWorld MT50avg success rate (%): 47.9 -> 85.0Improved
add ALAM latent transitions co-generated with actionspi0 · LIBEROavg success rate (%): 94.2 -> 98.1Improved
add ALAM versus JALA on LIBERO-Longpi0 · LIBERO-Longsuccess rate (%): 96.0 -> 94.4Hurt
add algebraic composition and reversal losses to latent pretraining (ablation)ALAM encoder · MetaWorld MT50avg success rate (%): 70.7 -> 85.0Improved
add reversal loss alone (ablation)ALAM encoder · MetaWorld MT50avg success rate (%): 71.6 -> 85.0Improved
add additivity loss alone (ablation)ALAM encoder · MetaWorld MT50avg success rate (%): 74.1 -> 85.0Improved
keep VQ bottleneck at pretraining (ablation)ALAM encoder · MetaWorld MT50avg success rate (%): 70.8 -> 85.0Improved
replace unstructured LAM with ALAM latents under joint flow matching (ablation)pi0 + latent stream · MetaWorld MT50avg success rate (%): 65.0 -> 85.0Improved
replace unstructured LAM with ALAM latents under joint diffusion (ablation)pi0 + latent stream · MetaWorld MT50avg success rate (%): 38.3 -> 65.6Improved
add unstructured LAM latents with joint diffusion (ablation)pi0 · MetaWorld MT50avg success rate (%): 47.9 -> 38.3Hurt
switch from two-stage to joint flow matching (ablation)ALAM · MetaWorld MT50avg success rate (%): 68.9 -> 85.0Improved
freeze, block or shuffle the latent stream at test time (ablation)pi0 + ALAM · MetaWorld MT50avg success rate (%): 85.0 -> 75.3-76.4Hurt

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

Experimental setup

Builds on: latent action models

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy