AI Paper Atlas

Research map / Reinforcement learning / Visual and representation RL

EfficientFlow: Efficient Equivariant Flow Policy Learning for Embodied AI

Jianlei Chang, Ruofeng Mei, Wei Ke, Xiangyu Xu

AAAI 2026 · Reinforcement learning › Visual and representation RL · approach: Skill learning and diffusion policies · 1 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add acceleration regularization (FABO) (ablation)EfficientFlow · MimicGen 12 tasks, 100 demosaverage success rate (%): 39.3 -> 52.6Improved
add equivariant architecture (ablation)EfficientFlow · MimicGen 12 tasks, 100 demosaverage success rate (%): 37.7 -> 52.6Improved
use acceleration regularization instead of Consistency Flow Matching (ablation)EfficientFlow · MimicGen 12 tasks, 100 demosaverage success rate (%): 40.4 -> 52.6Improved
use acceleration regularization instead of MeanFlow (ablation)EfficientFlow · MimicGen 12 tasks, 100 demosaverage success rate (%): 46.3 -> 52.6Improved
add acceleration regularization (ablation)EfficientFlow · MimicGen 12 tasks, 100 demosavg epochs to 50% of max success: 35.0 -> 31.7Improved
replace EquiDiff 100-step diffusion with 1-step flow policyEquiDiff · MimicGen 12 tasksinference time (ms): 685.92 -> 12.22Improved
replace EquiDiff with EfficientFlow (1 NFE)EquiDiff · MimicGen 12 tasks, 100 demosaverage success rate (%): 53.77 -> 52.61No clear effect
replace EquiDiff with EfficientFlow (5 NFE)EquiDiff · MimicGen 12 tasks, 1000 demosaverage success rate (%): 79.69 -> 81.99Improved
increase NFE from 1 to 5EfficientFlow · MimicGen 12 tasks, 1000 demosaverage success rate (%): 75.25 -> 81.99Improved
increase NFE from 1 to 5EfficientFlow · MimicGen 12 tasks, 100 demosaverage success rate (%): 52.61 -> 54.18Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy