AI Paper Atlas

Research map / Reinforcement learning / Offline RL

Look Beneath the Surface: Exploiting Fundamental Symmetry for Sample-Efficient Offline RL

Peng Cheng, Xianyuan Zhan, Zhihao Wu, Wenjia Zhang, Youfang Lin, Shou cheng Song, Han Wang, Li Jiang

NeurIPS 2023 · Reinforcement learning › Offline RL · approach: Simple practical offline RL · 2 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
use TSRL instead of DOGE on 10k datasetsTSRL · Hopper-medium 10knormalized score: 44.2 ± 10.2 -> 62.0±3.7Improved
use TSRL instead of best baseline on 10k Halfcheetah-expertTSRL · Halfcheetah-expert 10knormalized score: 4.2±0.94 -> 40.6±24.4Improved
use TSRL on full-size Hopper-medium vs DOGETSRL · Hopper-medium full 1Mnormalized score: 98.6 ± 2.1 -> 86.7±8.7Hurt
add T-symmetry data augmentation vs Gaussian noise (S4RL-N)TSRL · Walker2d-expert 10knormalized score: 56.5 ± 26.7 -> 102.2 ± 11.3Improved
replace TDM-based augmentation with CABITSRL · Hopper-expert 10knormalized score: 34.6 ± 24.4 -> 82.7 ± 21.9Improved
weaken TDM regularization (loss weights 0.01) (ablation)TSRL · Hopper-medium 10knormalized score: 62.0±3.7 -> 50.7±13.6Hurt
strengthen TDM regularization (loss weights 1) (ablation)TSRL · Hopper-medium 10knormalized score: 62.0±3.7 -> 43.6±14.3Hurt
weaken TDM regularization on Halfcheetah-medium (ablation)TSRL · Halfcheetah-medium 10knormalized score: 38.4±3.1 -> 39.1±3.6No clear effect
weaken TDM regularization on Walker2d-medium-expert (ablation)TSRL · Walker2d-medium-expert 10knormalized score: 46.4±17.4 -> 25.9±20.7Hurt
remove ODE structure from TDM (TDM-no-ODE) (ablation)TSRL · 10k MuJoCo datasetsnormalized scoreHurt
use T-symmetry regularized TDM representation instead of AE or SimSiam (ablation)TSRL · 10k MuJoCo datasetsnormalized scoreImproved
remove T-symmetry representation, policy constraint or augmentation (no-R, no-P, no-A) (ablation)TSRL · Halfcheetah-medium and medium-expert 10knormalized scoreHurt

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