AI Paper Atlas

Research map / Reinforcement learning / Visual and representation RL

CARP: Visuomotor Policy Learning via Coarse-to-Fine Autoregressive Prediction

Zhefei Gong, Pengxiang Ding, Shangke Lyu, Siteng Huang, Mingyang Sun, Wei Zhao, Zhaoxin Fan, Donglin Wang

ICCV 2025 · Reinforcement learning › Visual and representation RL · approach: Skill learning and diffusion policies · 28 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
replace Diffusion Policy with CARP next-scale autoregressive predictionDiffusion Policy · Robomimic state-based Squareinference time for 400 actions (s): 35.21 -> 3.07Improved
replace next-token autoregressive BET with CARPBET · Robomimic state-based Squaresuccess rate: 0.54 -> 0.98Improved
use CARP instead of diffusion policy on image-based SquareCARP · Robomimic image-based Squaresuccess rate: 0.92 -> 0.88Hurt
use CARP instead of SDP in multi-task visual settingSDP · MimicGen 8 tasksaverage success rate: 0.78 -> 0.85Improved
use CARP instead of SDP on Hammer CleanupSDP · MimicGen Hammer Cleanupsuccess rate: 1.00 -> 0.98Hurt
add EMA during training (ablation)CARP · training stability and performancesuccess rate: 4-5% improvementImproved
increase number of scales K above 4 (ablation)CARP · Can, Square, Kitchensuccess rateNo clear effect
use fewer scales (K below 4) (ablation)CARP · Can, Square, Kitchensuccess rateHurt
use cosine similarity instead of Euclidean distance in quantizer (ablation)CARP tokenizer · training stabilitytraining stabilityImproved
use separate VQ-VAE per action dimension instead of one joint VQ-VAE (ablation)CARP tokenizer · training stability and token interpretabilitystabilityImproved
use CARP instead of Diffusion Policy on real robots (RTX 2060)Diffusion Policy · real-world Cup and Bowlinference speed and success rate: about 8x faster, up to 10% higher successImproved
use CARP instead of VQ-BeT on Kitchen p4VQ-BeT · Franka Kitchenp4 success rate: 0.71 -> 0.98Improved

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