AI Paper Atlas

Research map / Reinforcement learning / Offline RL

$\text{DT}^\text{2}$: Decision-Targeted Digital Twins

Harry Amad, Mihaela van der Schaar

ICML 2026 · Reinforcement learning › Offline RL · approach: Diffusion models for offline decision making

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
train with decision-targeted ranking loss instead of MSEbottlenecked MLP digital twin · Limited neural capacity toyranking success rate: 20% -> 100%Improved
train with decision-targeted ranking loss instead of MSEMLP digital twin with partial observability · Partially observed toyranking accuracy: 65% -> 100%Improved
add ranking loss to NLL simulation lossexpressive DT architectures · Six control environmentsaverage regret: 4.08 -> 1.88Improved
add ranking loss to NLL simulation lossexpressive DT architectures · Six control environmentsaverage Spearman: 0.45 -> 0.66Improved
add ranking loss to NLL simulation lossexpressive DT architectures · Six control environmentstest transition MSE: 1.21 -> 1.41Hurt
add ranking loss to NLL simulation lossResNet DT · Cancerregret: 59.56 -> 26.96Improved
add ranking loss to NLL simulation lossResNet DT · CancerSpearman: 0.10 -> 0.64Improved
add ranking loss to NLL simulation lossResNet DT · Cancer (unseen policies)regret: 95.88 -> 49.27Improved
add ranking loss to NLL simulation lossResNet DT · Cancer (unseen policies)Spearman: 0.40 -> 0.50Improved
use DT2 instead of FQE ranking directlyDT2 · Six control envsregret, Spearman: Ant regret 0 vs 1.64; Walker regret 1.18 vs 0.21Mixed
compare MBRL pessimism methods MOReL, MOPO, ROMI with NLL DTsoffline MBRL baselines · Six control envsregret, SpearmanNo clear effect
increase ranking weight lambda from 0 up to 0.75 (ablation)DT2 ResNet · Six control envsSpearman and test MSEMixed

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