SteerAct: Unlocking Action Models through Few-Shot Temporal Steering

Key points

• Improving pretrained robot action models at test time from only a few rollouts remains challenging.1
• Activation steering offers a lightweight approach by modifying feature activations while keeping model weights fixed.1
• A pretrained sparse autoencoder provides candidate features, and our temporal descriptor summarizes the distribution of local activation-order patterns for each feature, enabling comparisons across rollouts without requiring timestep alignment.1
• Using only 24 outcome-labeled rollouts per task, SteerAct improves task success by approximately 10% relative to unsteered models in evaluations spanning $π{0.5}$, Fast-WAM, RoboTwin 2.0, DexJoCo, and real-world manipulation, without updating action-model weights.1

Sources (1)

• [1]SteerAct: Unlocking Action Models through Few-Shot Temporal Steering
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 19, 01:20 AM

  We introduce SteerAct, a few-shot temporal steering framework that uses a shared temporal descriptor of sparse-feature dynamics for offline feature selection and online control.

  Improving pretrained robot action models at test time from only a few rollouts remains challenging.

Extractive summary: sentences quoted from the sources.