AI Paper Atlas

Research map / Reinforcement learning / Imitation learning

Dense Policy: Bidirectional Autoregressive Learning of Actions

Yue Su, Xinyu Zhan, Hongjie Fang, Han Xue, Hao-Shu Fang, Yong-Lu Li, Cewu Lu, Lixin Yang

ICCV 2025 · Reinforcement learning › Imitation learning · approach: Imitation combined with RL · 15 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 action head with Dense Policy action head (3D)DP3 · Adroit/DexArt/MetaWorldsuccess rate (%): 53 -> 72Improved
replace diffusion action head with Dense Policy action head (2D)Diffusion Policy · Adroit/DexArt/MetaWorldsuccess rate (%): 25 -> 52Improved
use bidirectional prediction instead of next-token or next-chunk autoregression (ablation)Dense Policy · Door, Bin Picking, Shelf Place, Box Closelearning efficiency and success rateImproved
3D Dense Policy vs RISE on Pour Balls complete transferRISE · Pour Balls (real, 3D)complete (%): 25 -> 60Improved
3D Dense Policy vs RISE on Pour Balls poured at least one ballRISE · Pour Balls (real, 3D)poured (%): 95 -> 85Hurt
3D Dense Policy vs RISE on Flower ArrangementRISE · Flower Arrangement (real, 3D)success rate (%): 50 -> 70Improved
2D Dense Policy vs Diffusion Policy on Put BreadDiffusion Policy · Put Bread (real, 2D)success rate (%): 40 -> 55Improved
2D Dense Policy vs Diffusion Policy on Open DrawerDiffusion Policy · Open Drawer (real, 2D)success rate (%): 20 -> 20No clear effect
2D Dense Policy vs Diffusion Policy on Pour BallsDiffusion Policy · Pour Balls (real, 2D)balls poured out of 10: 2.35 -> 3.30Improved
2D Dense Policy vs ACT on Put BreadACT · Put Bread (real, 2D)success rate (%): 35 -> 55Improved
Dense Policy action head vs Diffusion Policy action head, inference speedDiffusion Policy · action head inferenceinference time: nearly ten times fasterImproved

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