AI Paper Atlas

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

The Cost of Information: Phase Transitions in Contextual Bandits with Paid Observations

Xueping Gong, Jiheng Zhang

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
replace enumerated policy class (Algorithm 1) with function-approximation oracle (Algorithm 3)paid-observation contextual bandit · Run time (T=40000, |A|=10, d=8)execution time (s): 94.0 -> 2.0Improved
increase policy class size from 10 to 40 (ablation)Algorithm 1 · Run time (T=40000, |A|=10, d=8)execution time (s): 33.2 s to 94.0 sHurt
make observation-cost-aware (Algorithm 3 vs linUCB)linUCB · d=10, |A|=5, T=10000, c=0.5final average cumulative regret: 5003.32 -> 2899.75Improved
Algorithm 3 versus Tucker et al. 2023cost-aware baseline · d=10, |A|=5, T=10000, c=0.5final average cumulative regret: 3697.13 -> 2899.75Improved
larger policy space (ablation)Algorithm 1 · synthetic gridcumulative regretHurt
more actions (ablation)Algorithm 1 · synthetic gridcumulative regretHurt
increase context dimension d from 2 to 8 (ablation)Algorithm 1 · synthetic gridcumulative regretNo clear effect

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