Steering Robustness into World Action Models via Mechanistic Interpretability and Optimal Control

• URL: http://arxiv.org/abs/2607.14943v1
• Date: Thu, 16 Jul 2026 12:52:49 GMT
• Title: Steering Robustness into World Action Models via Mechanistic Interpretability and Optimal Control
• Abstract summary: We study how robustness-relevant perturbations are represented in World Action Models (WAMs) activation space.<n> Comparing activations across successful and unsuccessful rollouts, we find some WAM architectures exhibit low-dimensional linear separability for robustness-critical features.<n>This motivates the use of contrastive activation directions for training-free WAM steering.<n>We also show that local linearity in WAM activation dynamics enables efficient feedback steering via model-based optimal control.
• Score: 3.3394856680250284
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: World Action Models (WAMs) enable semantically- and physically-informed control but are brittle under distribution shift. In this work, we use mechanistic interpretability to study how robustness-relevant perturbations are represented in WAM activation space. Comparing activations across successful and unsuccessful rollouts, we find some WAM architectures exhibit low-dimensional linear separability for robustness-critical features, while others do not. This motivates the use of contrastive activation directions for training-free WAM steering. We also show that local linearity in WAM activation dynamics enables efficient feedback steering via model-based optimal control, yielding World-Action Linear Quadratic Regulator (WA-LQR), a minimally-invasive reduced-order LQR controller. Via mechanistic evaluations, we predict strong steerability in the Cosmos-Policy and DiT4DiT models but weak steerability in LingBot-VA, consistent with steering intervention results. On Cosmos-Policy and DiT4DiT, WA-LQR generalizes contrastive directions to new tasks and improves robustness to camera, gripper, and visual-noise perturbations over unsteered and prompt steering baselines.

Related papers

• A Neural Network Based Teleoperation for Remote Controlled Vehicles [98.38733381483934]
  Direct teleoperation of vehicles faces critical technical bottlenecks.<n> communication latency and the operator's inability to physically perceive unmodeled environmental disturbances.<n>We propose a tailored unilateral teleoperation framework.
  arXiv  Detail & Related papers  (2026-08-11T01:51:08Z)
• Activation Steering Induces Emergent Misalignment: A More Comprehensive Evaluation [74.17379276939599]
  We show that activation steering can induce broad misalignment, even in the recent Qwen-3.5 series.<n>We characterize properties of AS-induced EM by analyzing key steering-specific factors, including steering magnitude, the low-rank structure of the steering subspace, and the number of epochs during steering-vector construction.
  arXiv  Detail & Related papers  (2026-06-07T15:34:59Z)
• Principled Steering via Null-space Projection for Jailbreak Defense in Vision-Language Models [71.11171576590134]
  Vision-language models can be easily induced by visual jailbreak attacks to generate harmful content.<n>We propose NullSteer, a null-space projected activation defense framework.<n>We show that NullSteer significantly reduces harmful outputs under various jailbreak attacks.
  arXiv  Detail & Related papers  (2026-03-23T15:23:23Z)
• Data-Driven Physics Embedded Dynamics with Predictive Control and Reinforcement Learning for Quadrupeds [2.85987101794478]
  State of the art quadrupedal approaches integrate Model Predictive Control (MPC) with Reinforcement Learning (RL)<n>We address these issues by integrating Lagrangian Neural Networks (LNNs) into an RL MPC framework.<n>We show improved sample efficiency, reduced long-horizon error, and faster real time planning compared to unstructured neural dynamics.
  arXiv  Detail & Related papers  (2026-03-15T11:52:14Z)
• ResWM: Residual-Action World Model for Visual RL [0.06999740786886534]
  Learning predictive world models from raw visual observations is a central challenge in reinforcement learning (RL)<n>We introduce the Residual-Action World Model (ResWM), a new framework that reformulates the control variable from absolute actions to residual actions.<n>ResWM produces more stable and energy-efficient action trajectories, a property critical for robotic systems deployed in real-world environments.
  arXiv  Detail & Related papers  (2026-03-11T11:27:08Z)
• Why Steering Works: Toward a Unified View of Language Model Parameter Dynamics [81.80010043113445]
  Local weight fine-tuning, LoRA-based adaptation, and activation-based interventions are studied in isolation.<n>We present a unified view that frames these interventions as dynamic weight updates induced by a control signal.<n>Across methods, we observe a consistent trade-off between preference and utility: stronger control increases preference while predictably reducing utility.
  arXiv  Detail & Related papers  (2026-02-02T17:04:36Z)
• Efficiently Learning Robust Torque-based Locomotion Through Reinforcement with Model-Based Supervision [14.246416873754905]
  We propose a control framework that integrates model-based bipedal locomotion with residual reinforcement learning.<n>Our method demonstrates improved robustness and generalization across a range of randomized conditions.
  arXiv  Detail & Related papers  (2026-01-22T16:56:52Z)
• Mechanistic interpretability for steering vision-language-action models [0.23371356738437823]
  Vision-Language-Action (VLA) models are a promising path to realizing generalist embodied agents.<n>We introduce the first framework for interpreting and steering VLAs via their internal representations.<n>We introduce a general-purpose activation steering method that modulates behavior in real time, without fine-tuning, reward signals, or environment interaction.
  arXiv  Detail & Related papers  (2025-08-30T03:01:57Z)
• Model-Based Reinforcement Learning with Isolated Imaginations [61.67183143982074]
  We propose Iso-Dream++, a model-based reinforcement learning approach. We perform policy optimization based on the decoupled latent imaginations. This enables long-horizon visuomotor control tasks to benefit from isolating mixed dynamics sources in the wild.
  arXiv  Detail & Related papers  (2023-03-27T02:55:56Z)
• Incorporating Recurrent Reinforcement Learning into Model Predictive Control for Adaptive Control in Autonomous Driving [11.67417895998434]
  Model Predictive Control (MPC) is attracting tremendous attention in the autonomous driving task as a powerful control technique. In this paper, we reformulate the problem as a Partially Observed Markov Decision Process (POMDP) We then learn a recurrent policy continually adapting the parameters of the dynamics model via Recurrent Reinforcement Learning (RRL) for optimal and adaptive control.
  arXiv  Detail & Related papers  (2023-01-30T22:11:07Z)
• Isolating and Leveraging Controllable and Noncontrollable Visual Dynamics in World Models [65.97707691164558]
  We present Iso-Dream, which improves the Dream-to-Control framework in two aspects. First, by optimizing inverse dynamics, we encourage world model to learn controllable and noncontrollable sources. Second, we optimize the behavior of the agent on the decoupled latent imaginations of the world model.
  arXiv  Detail & Related papers  (2022-05-27T08:07:39Z)

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.