Research map / Probabilistic and scientific modeling / Uncertainty and Gaussian processes
Balancing Risk and Reward: A Batched-Bandit Strategy for Automated Phased Release
Yufan Li, Jialiang Mao, Iavor Bojinov
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 changed | Where | Measured | Effect |
|---|---|---|---|
| larger budget B and larger risk tolerance delta (ablation) | RRC ramp-up algorithm · Simulated PTE/NTE/NPTE ramp schedules | treated units / ramp-up speed | Improved |
| inadequate budget with low ruin tolerance (ablation) | RRC ramp-up algorithm · NPTE scenario | reaching 50% treatment (mt=250) | Hurt |
| reserve budget for later stages by lowering bt or Delta_t in early stages (ablation) | RRC ramp-up algorithm · NPTE scenario | ramp-up speed | Improved |
| use proposed algorithm instead of Thompson-sampling Bayesian bandit | phased release · PNTE simulation | ramp schedule respecting budget | Improved |
| vary Thompson-sampling tuning parameter c (ablation) | Thompson-sampling bandit · NPTE and LinkedIn | ramp aggressiveness | Mixed |
| estimate variance with Eq. (6) instead of known variance (ablation) | RRC ramp-up algorithm · Simulated ramp schedules | ramp schedule | No clear effect |
| non-Gaussian outcomes (correlated, Bernoulli, fat-tailed) instead of Gaussian (ablation) | RRC ramp-up algorithm · Budget-spent simulation | actual risk of ruin: 1.22% -> 1.52% | No clear effect |
| treatment effect decreasing over stages (ablation) | RRC ramp-up algorithm · Budget-spent simulation (dec) | actual risk of ruin: 1.22% -> 18.28% | Hurt |
Closest papers by meaning
- A Bandit Approach to Sequential Experimental Design with False Discovery Control (NeurIPS 2018 · similarity 0.86)
- Statistical Inference on Multi-armed Bandits with Delayed Feedback (ICML 2023 · similarity 0.86)
- Myopic Posterior Sampling for Adaptive Goal Oriented Design of Experiments (ICML 2019 · similarity 0.86)
- Strategic A/B testing via Maximum Probability-driven Two-armed Bandit (ICML 2025 · similarity 0.86)
- Simulation-Based Inference for Adaptive Experiments (NeurIPS 2025 · similarity 0.85)
- Noise-Adaptive Confidence Sets for Linear Bandits and Application to Bayesian Optimization (ICML 2024 · similarity 0.84)
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.
