AI Paper Atlas

Research map / Bandits and online learning / Contextual bandits and Thompson sampling

Adaptively Grouped Contextual Bandits for Heterogeneous Human-AI Decision Making with Conformal Prediction Sets

Yanchen Wu, Bo Li

ICML 2026 · Bandits and online learning › Contextual bandits and Thompson sampling · approach: Oracle-efficient contextual bandits

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
add adaptive grouping with counterfactual updates to Sequential HalvingSH · Synthetic (100 arms, 5-dim), binarysimple regret: 0.381 -> 0.298Improved
add adaptive grouping with counterfactual updates to Sequential HalvingSH · Synthetic (100 arms, 5-dim), continuoussimple regret: 0.211 -> 0.111Improved
add adaptive grouping with counterfactual updates to Sequential HalvingSH · Synthetic (100 arms, 10-dim), binarysimple regret: 0.421 -> 0.307Improved
add adaptive grouping with counterfactual updates to Sequential HalvingSH · Synthetic (100 arms, 10-dim), continuoussimple regret: 0.251 -> 0.113Improved
add adaptive grouping with counterfactual updates to Sequential HalvingSH · ImageNet16H-PS, continuoussimple regret: 0.205 -> 0.101Improved
use AGCB-SH instead of BOED in the small-arm regimeAGCB-SH · Synthetic (10 arms, 5-dim), binarysimple regret: 0.193 vs 0.190No clear effect
raise context dimension from 5 to 10AGCB-SH · Synthetic (100 arms), binarysimple regret: 0.298 -> 0.307No clear effect
add AGCB grouping on top of UCB with counterfactual updatesUCB (with CF) · ImageNet16H-PS, binarycumulative regret: 93.97 -> 75.93Improved
add AGCB grouping on top of UCB with counterfactual updatesUCB (with CF) · ImageNet16H-PS, continuouscumulative regret: 109.86 -> 107.24Improved
add AGCB grouping on top of SE with counterfactual updatesSE (with CF) · ImageNet16H-PS, binarycumulative regret: 122.16 -> 93.87Improved
add AGCB grouping on top of SE with counterfactual updatesSE (with CF) · ImageNet16H-PS, continuouscumulative regret: 143.87 -> 106.23Improved
add counterfactual updates to UCBUCB · ImageNet16H-PS, binarycumulative regret: 150.23 -> 93.97Improved

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