AI Paper Atlas

Research map / Reinforcement learning / Offline RL

Cross-Embodiment Offline Reinforcement Learning for Heterogeneous Robot Datasets

Haruki Abe, Takayuki Osa, Yusuke Mukuta, Tatsuya Harada

ICLR 2026 · Reinforcement learning › Offline RL · approach: Simple practical offline RL · 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
use IQL instead of BCcross-embodiment training · 70% Suboptimal Forwardfinal return: 30.52 -> 36.62Improved
use IQL instead of BCcross-embodiment training · 70% Suboptimal Backwardfinal return: 41.42 -> 38.69Hurt
use IQL instead of BCcross-embodiment training · Expert Forward/Backwardfinal return: 63.31 -> 63.39No clear effect
add embodiment grouping to IQLIQL · 70% Suboptimal Forwardfinal return: 36.62 -> 51.19Improved
add embodiment grouping to IQLIQL · Mean over six datasetsfinal return: 52.05 -> 57.29Improved
add embodiment grouping to IQL vs adding PCGrad or SELIQL · 70% Suboptimal datasetsrelative improvement over IQL: 7.15% for PCGrad, 18.33% for SEL and 33.99% for EGImproved
add embodiment grouping to TD3+BCTD3+BC · 70% Suboptimal datasetsfinal return: +19.5% on average; mean 46.43 -> 51.39Improved
add embodiment grouping to BCBC · 70% Suboptimal splitsfinal return: +26.3%Improved
add embodiment grouping to IQLIQL · Expert Forwardfinal return: 63.39 -> 63.52No clear effect
group robots by FGW embodiment distance vs random grouping (ablation)IQL+EG · 70% Suboptimal Forwardfinal return: 38.73 -> 51.98Improved
group robots by FGW embodiment distance vs heuristic split (ablation)IQL+EG · 70% Suboptimal Forwardfinal return: 34.45 -> 51.98Improved
add embodiment grouping with compute-normalized IQL baseline (ablation)IQL · 70% Suboptimal Forwardfinal return: 44.20 -> 51.98Improved

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