cs.ROJul 9, 2026

FlowDAgger: Human-in-the-Loop Adaptation of Generative Robot Policies in Latent Space

Authors: Michael MurrayDaphne ChenSimran BagariaDean FortierTess HellebrekersGalen MullinsHarshavardhan GajarlaOier Mees+2 more

Organizations: 1Microsoft Research · University of Washington · 2Microsoft · 3ETH Zurich

Abstract

Pretrained generative robot policies based on flow matching and diffusion have achieved impressive results across a wide range of manipulation tasks. Yet real-world deployments routinely expose failure modes outside the pretraining distribution. Closing these gaps typically requires large-scale data collection or online reinforcement learning on physical hardware, which is impractical for rapid and safe adaptation. We present FlowDAgger, a sample- and compute-efficient method for adapting frozen generative robot policies from human interventions in latent space. Our key idea is action inversion: each human expert action is mapped to the noise that would have produced it under the frozen base policy, using reverse-time integration followed by local refinement. The resulting inverted noise provides supervision for a lightweight latent policy that steers the base model at deployment time, enabling rapid skill acquisition while preserving its behavioral priors. We evaluate FlowDAgger in simulation and on real-world bimanual and single-arm manipulation, adapting both action-head VLAs and world-action models from a handful of interventions. FlowDAgger outperforms supervised fine-tuning and latent-space RL baselines and preserves pretrained skills on held-out tasks, offering a practical path for adapting robot foundation models in the real world. Website: https://microsoft.github.io/FlowDAgger

Explore similar work

CardsList
  1. Flow-based Policy Adaptation without Policy Updates

    Jun 4, 2026Luzhe Sun, Jingtian Ji, Haoran Chen +2Skewed Expert Demand