Skip to main content
Search

# One-Step Flow Policy

1 related articles

Subscribe to updates
  1. arXiv

    OMAF replaces diffusion policies with one-step flow policies, reaching up to 3.4x returns and 10.5x sample efficiency on multi-agent tasks

    The work proposes OMAF, an online multi-agent reinforcement learning framework that replaces iterative-sampling diffusion policies with a Transformer-based one-step flow policy, combined with an approximate path score surrogate and a joint optimization scheme coupling softmax Q-value estimation with a joint flow policy objective, achieving up to 3.4x higher returns and 10.5x sample efficiency improvement across 10 standard tasks from MPE and MAMuJoCo.

You’re all caught up