Contact-Anchored Policies: Contact Conditioning Creates Strong Robot Utility Models

TL;DR

This paper introduces Contact-Anchored Policies (CAP), a modular, contact-based approach for robot manipulation that improves generalization and efficiency through simulation-based iteration and minimal demonstration data.

Contribution

The work presents a novel contact-anchored policy framework that replaces language conditioning with physical contact points and uses a simulation iteration cycle for rapid refinement.

Findings

CAP generalizes to new environments and embodiments.

Achieves strong zero-shot performance, outperforming state-of-the-art methods by 56%.

Uses only 23 hours of demonstration data.

Abstract

The prevalent paradigm in robot learning attempts to generalize across environments, embodiments, and tasks with language prompts at runtime. A fundamental tension limits this approach: language is often too abstract to guide the concrete physical understanding required for robust manipulation. In this work, we introduce Contact-Anchored Policies (CAP), which replace language conditioning with points of physical contact in space. Simultaneously, we structure CAP as a library of modular utility models rather than a monolithic generalist policy. This factorization allows us to implement a real-to-sim iteration cycle: we build EgoGym, a lightweight simulation benchmark, to rapidly identify failure modes and refine our models and datasets prior to real-world deployment. We show that by conditioning on contact and iterating via simulation, CAP generalizes to novel environments and embodiments out of the box on three fundamental manipulation skills while using only 23 hours of demonstration data, and outperforms large, state-of-the-art VLAs in zero-shot evaluations by 56%. All model checkpoints, codebase, hardware, simulation, and datasets will be open-sourced. Project page: https://cap-policy.github.io/