DREAMSTEER: Latent World Models Can Steer VLA Policies During Deployment Without Any Finetuning

• URL: http://arxiv.org/abs/2607.02865v1
• Date: Fri, 03 Jul 2026 02:08:50 GMT
• Title: DREAMSTEER: Latent World Models Can Steer VLA Policies During Deployment Without Any Finetuning
• Abstract summary: DREAMSTEER is a deployment-time steering framework for pretrained vision-language-action (VLA) policies.<n>Itleverages a latent world model and a value model to steer pretrained VLA policies.<n>It improves task success rate from 23.75% to 66.25% and instruction-following accuracy from 38.75% to 56.25% over the base VLA policy.
• Score: 19.23819338884138
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Pretrained vision-language-action (VLA) policies show promising zero-shot generalization, but often fail under deployment-time distribution shift, leading to decreased robustness and inconsistent instruction following. While prior work commonly tackles this by finetuning on in-distribution data, it assumes demonstrations collected on tasks in the target environment. In this work, we propose DREAMSTEER, a deployment-time steering framework for pretrained VLAs without any finetuning or parameter modifications. The key insight in DREAMSTEER is to leverage a latent world model and a value model to steer pretrained VLA policies. During deployment, DREAMSTEER samples candidate action chunks from a VLA policy and predefined motion primitives, imagines their outcomes using an action-conditioned latent world model, and ranks the imagined trajectories with a language-conditioned value model. Across four real-world manipulation benchmarks with unseen objects, DREAMSTEER improves task success rate from 23.75% to 66.25% and instruction-following accuracy from 38.75% to 56.25% over the base VLA policy.

Related papers

• World Pilot: Steering Vision-Language-Action Models with World-Action Priors [21.867210989881016]
  Vision-Language-Action (VLA) models inherit semantic grounding from large-scale pretraining.<n>World Pilot augments the policy with priors from a World-Action Model (WAM)<n>World Pilot attains a state-of-the-art Total success rate of 84.7% on the LIBERO-Plus zero-shot OOD benchmark.
  arXiv  Detail & Related papers  (2026-06-10T17:59:08Z)
• Grounded World Model for Semantically Generalizable Planning [94.53923128709965]
  We learn a Grounded World Model (GWM) in a vision-language-aligned latent space.<n>Each proposed action is scored based on how close its future outcome is to the task instruction.<n>On the proposed WISER benchmark, GWM-MPC achieves a 87% success rate on the test set.
  arXiv  Detail & Related papers  (2026-04-13T17:25:41Z)
• Enhancing Policy Learning with World-Action Model [4.941630596191806]
  World-Action Model (WAM) is an action-regularized world model that jointly reasons over future visual observations and the actions that drive state transitions.<n>WAM incorporates an inverse dynamics objective into DreamerV2 that predicts actions from latent state transitions.<n>We evaluate WAM on enhancing policy learning across eight manipulation tasks from the CALVIN benchmark.
  arXiv  Detail & Related papers  (2026-03-30T19:56:05Z)
• Scaling World Model for Hierarchical Manipulation Policies [61.736772957803026]
  Vision-Language-Action (VLA) models are promising for generalist robot manipulation but brittle in out-of-distribution settings.<n>We introduce a hierarchical Vision-Language-Action framework our that leverages the generalization of large-scale pre-trained world model.<n>We validate both visual goal synthesis and our hierarchical VLA policies in massive out-of-distribution scenarios.
  arXiv  Detail & Related papers  (2026-02-11T16:12:33Z)
• Reshaping Action Error Distributions for Reliable Vision-Language-Action Models [69.38615670891038]
  In robotic manipulation, vision-language-action (VLA) models have emerged as a promising paradigm for learning generalizable and scalable robot policies.<n>We focus on continuous-action VLA models and move beyond conventional MSE-based regression by reshaping action error distributions during training.<n>We evaluate our approaches across standard, few-shot, and noisy settings on multiple representative VLA architectures.
  arXiv  Detail & Related papers  (2026-02-04T05:37:09Z)
• Align-Then-stEer: Adapting the Vision-Language Action Models through Unified Latent Guidance [63.33213516925946]
  We introduce textbfAlign-Then-stEer (textttATE), a novel, data-efficient, and plug-and-play adaptation framework.<n>Our work presents a general and lightweight solution that greatly enhances the practicality of deploying VLA models to new robotic platforms and tasks.
  arXiv  Detail & Related papers  (2025-09-02T07:51:59Z)
• Fine-Tuning Vision-Language-Action Models: Optimizing Speed and Success [100.226572152954]
  We present an optimized fine-tuning recipe for vision-language-action models (VLAs)<n>Our recipe boosts OpenVLA's average success rate across four task suites from 76.5% to 97.1% while increasing action generation throughput by 26$times$.<n>In real-world evaluations, our fine-tuning recipe enables OpenVLA to successfully execute dexterous, high-frequency control tasks on a bimanual ALOHA robot.
  arXiv  Detail & Related papers  (2025-02-27T00:30:29Z)
• GRAPE: Generalizing Robot Policy via Preference Alignment [58.419992317452376]
  We present GRAPE: Generalizing Robot Policy via Preference Alignment.<n>We show GRAPE increases success rates on in-domain and unseen manipulation tasks by 51.79% and 58.20%, respectively.<n> GRAPE can be aligned with various objectives, such as safety and efficiency, reducing collision rates by 37.44% and rollout step-length by 11.15%, respectively.
  arXiv  Detail & Related papers  (2024-11-28T18:30:10Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.