# Inference-time Policy Steering via Vision and Touch

arXiv: https://arxiv.org/abs/2606.14981 · HTML: https://mlsanity.com/paper/2606.14981/

## Key facts

- Cited by 2 papers in the map (data through 14 Oct 2026).
- Builds on 20 of its 56 references that are in the map.
- First posted on arXiv 12 Jun 2026.

## Builds on (most cited first, 20 of 20)

- [π0: A Vision-Language-Action Flow Model for General Robot Control](https://mlsanity.com/paper/2410.24164/) (2024)
- [SDEdit: Guided Image Synthesis and Editing with Stochastic Differential Equations](https://mlsanity.com/paper/2108.01073/) (2021)
- [Dream to Control: Learning Behaviors by Latent Imagination](https://mlsanity.com/paper/1912.01603/) (2019)
- [DINOv3](https://mlsanity.com/paper/2508.10104/) (2025)
- [V-JEPA 2: Self-Supervised Video Models Enable Understanding, Prediction and Planning](https://mlsanity.com/paper/2506.09985/) (2025)
- [Diffusion Transformers with Representation Autoencoders](https://mlsanity.com/paper/2510.11690/) (2025)
- [Steering Your Diffusion Policy with Latent Space Reinforcement Learning](https://mlsanity.com/paper/2506.15799/) (2025)
- [Ctrl-World: A Controllable Generative World Model for Robot Manipulation](https://mlsanity.com/paper/2510.10125/) (2025)
- [Training Agents Inside of Scalable World Models](https://mlsanity.com/paper/2509.24527/) (2025)
- [DreamDojo: A Generalist Robot World Model from Large-Scale Human Videos](https://mlsanity.com/paper/2602.06949/) (2026)
- [Cosmos-Transfer1: Conditional World Generation with Adaptive Multimodal Control](https://mlsanity.com/paper/2503.14492/) (2025)
- [RoboMonkey: Scaling Test-Time Sampling and Verification for Vision-Language-Action Models](https://mlsanity.com/paper/2506.17811/) (2025)
- [Inference-Time Policy Steering through Human Interactions](https://mlsanity.com/paper/2411.16627/) (2024)
- [ViTacFormer: Learning Cross-Modal Representation for Visuo-Tactile Dexterous Manipulation](https://mlsanity.com/paper/2506.15953/) (2025)
- [OmniVTA: Visuo-Tactile World Modeling for Contact-Rich Robotic Manipulation](https://mlsanity.com/paper/2603.19201/) (2026)
- [Visuo-Tactile World Models](https://mlsanity.com/paper/2602.06001/) (2026)
- [VJEPA: Variational Joint Embedding Predictive Architectures as Probabilistic World Models](https://mlsanity.com/paper/2601.14354/) (2026)
- [VTAM: Video-Tactile-Action Models for Complex Physical Interaction Beyond VLAs](https://mlsanity.com/paper/2603.23481/) (2026)
- [Multi-Modal Manipulation via Multi-Modal Policy Consensus](https://mlsanity.com/paper/2509.23468/) (2025)
- [TouchGuide: Inference-Time Steering of Visuomotor Policies via Touch Guidance](https://mlsanity.com/paper/2601.20239/) (2026)

## Built on by (most cited first, 2 of 2)

- [Learning Physical Interaction: A Survey of Tactile- and Force-aware Robot Learning](https://mlsanity.com/paper/2608.07558/) (2026)
- [Correcting WHERE, Preserving HOW: Compositional Generalization for Vision-Language-Action Models via Referential Guidance](https://mlsanity.com/paper/2609.38616/) (2026)

## Abstract (arXiv)

Inference-time steering adapts pre-trained generative robot policies during deployment by verifying candidate actions before execution. While prior methods typically perform this verification only with visual observations, vision alone is often insufficient for contact-rich manipulation, where success depends on both global task progress and subtle local interactions such as contact force. We introduce ViTaL, a visuo-tactile inference-time steering framework that formulates multimodal guidance as a bi-level optimization problem. At the high level, visual sampling-and-verification performs long-horizon mode selection, deciding what behavior the robot should execute. At the low level, tactile-guided diffusion editing refines the selected action sequence over a shorter horizon to satisfy local contact requirements. To support outcome-based steering, ViTaL learns a visuo-tactile latent world model and employs semantically aligned visual and tactile verifiers, including a novel text-conditioned tactile reward that scores predicted tactile futures directly in latent space. Across three real-world contact-rich manipulation tasks, ViTaL improves overall success by 51% over the base policy, outperforms unimodal steering by at least 33%, and exceeds naive multimodal fusion by at least 20%. Website: https://yilin-wu98.github.io/vital_website.