AI Paper Atlas

Research map / Reinforcement learning / Visual and representation RL

Diffusion Model is an Effective Planner and Data Synthesizer for Multi-Task Reinforcement Learning

Haoran He, Chenjia Bai, Kang Xu, Zhuoran Yang, Weinan Zhang, Dong Wang, Bin Zhao, Xuelong Li

NeurIPS 2023 · Reinforcement learning › Visual and representation RL · approach: In-context and adaptive RL · 12 citations (OpenAlex)

Read the paper

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
use GPT2 transformer backbone instead of U-Net (ablation)MTDiff-p · MT50-rand near-optimalsuccess rate (%): 55.67 -> 59.53Improved
use GPT2 transformer backbone instead of U-Net (ablation)MTDiff-p · MT50-rand sub-optimalsuccess rate (%): 47.42 -> 48.67Improved
replace one-hot task ID with trajectory prompt (ablation)MTDiff-p · MT50-rand near-optimalsuccess rate (%): 61.32 -> 59.53No clear effect
replace one-hot task ID with trajectory prompt (ablation)MTDiff-p · MT50-rand sub-optimalsuccess rate (%): 48.94 -> 48.67No clear effect
use diffusion planning with prompt instead of autoregressive PromptDTPrompt-conditioned GPT2 policy · Meta-World MT50-rand near-optimalsuccess rate (%): 45.68 -> 59.53Improved
use diffusion planning instead of PromptDTPrompt-conditioned GPT2 policy · Ant-diraverage score: 409.81 -> 602.17Improved
augment dataset with MTDiff-s synthetic dataTD3-BC · coffee-pushsuccess rate (%): 28.60 -> 74.67Improved
augment dataset with MTDiff-s synthetic dataTD3-BC · hand-insert unseensuccess rate (%): 30.60 -> 70.87Improved
augment dataset with MTDiff-s synthetic dataTD3-BC · sweep-intosuccess rate (%): 91.8 -> 92.87No clear effect
MTDiff-s synthetic data vs S4RL augmentationTD3-BC · bin-picking unseensuccess rate (%): 72.20 -> 55.73Hurt
train MTDiff-s on more tasks (10, 20, 30 vs single task) (ablation)MTDiff-s · coffee-pushsuccess rateImproved

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