# GUARD steers robot policies away from risky actions

> A new framework uses anomaly scores from a policy’s training data to guide robot actions. The authors report gains in simulated and real-world manipulation tasks.

Oossa · 2026-11-10 · https://oossa.com/en/guard-steers-robot-policies-away-from-risky-actions

Researchers introduced GUARD, a method for steering generative robot policies during execution without changing the pretrained policy. It combines task rewards with anomaly scores—signals that an action falls outside patterns supported by training data—and uses a learned Q-function to guide the policy’s action-generation process.

In four simulated dexterous-manipulation tasks, the authors report average success rates of 96% under normal conditions and 80% under perturbed dynamics, beating the strongest baseline by 21 and 20 percentage points. They also report success-rate gains of 40 points for in-hand screwdriver rotation and 25 points for reorientation in direct sim-to-real tests.

## The facts

- The paper reports tests across four Adroit dexterous-manipulation tasks.
- The arXiv submission history dates the first version to November 7, 2026.

## Why it matters

For robotics teams adapting generative policies to changing conditions, GUARD offers a way to use training-data anomalies as a warning signal during action selection; the paper reports promising results, but does not establish how well the method works beyond its tested tasks.

## Sources & references

1. [GUARD: Anomaly-Informed Q-Steering for Generative Robot Policies in Dexterous Manipulation](https://arxiv.org/abs/2611.10343) – arXiv cs.RO, 2026-11-10

Last updated: 2026-11-10