Research map / Bandits and online learning / Contextual bandits and Thompson sampling
IBCB: Efficient Inverse Batched Contextual Bandit for Behavioral Evolution History
Yi Xu, Weiran Shen, Jun Xu, Xiao Zhang, Ji-Rong Wen
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 |
|---|---|---|---|
| use IBCB instead of B-IRL | imitation of SBUCB expert · Synthetic (std=0) | BT-Fitness: 0.828 -> 0.972 | Improved |
| use IBCB instead of BaseSVC | imitation of SBUCB expert · Synthetic (std=0) | BT-Fitness: 0.791 -> 0.972 | Improved |
| use IBCB instead of B-ICB-S | imitation of SBTS expert · Synthetic (std=0) | BT-Fitness: 0.740 -> 0.963 | Improved |
| use IBCB instead of B-IRL for training speed | expert imitation · Synthetic (std=0) | train time (sec): 36.997 -> 0.815 | Improved |
| use IBCB instead of BaseSVC for training speed | expert imitation · ML-100K (SBUCB) | train time (sec): 162.767 -> 4.036 | Improved |
| use IBCB instead of B-IRL for reward-parameter fit | SBUCB expert · Synthetic (std=0) | OL-Fitness: 0.868 vs 0.827 | Hurt |
| IBCB average reward versus expert SBUCB | IBCB · Synthetic (std=0) | BT-AR: 0.647 vs 0.641 | No clear effect |
| use IBCB instead of B-IRL under distribution shift | SBUCB expert · Synthetic OOD (std=0) | BT-Fitness: 0.802 -> 0.970 | Improved |
| use IBCB instead of B-IRL | SBUCB expert · ML-100K | BT-Fitness: 0.864 -> 0.925 | Improved |
| use IBCB instead of B-ICB-S | SBUCB expert · ML-100K | BT-Fitness: 0.767 -> 0.925 | Improved |
| use IBCB instead of BaseSVC for fairness | SBUCB top-2 expert · Synthetic fairness (std=0) | CFR: 2027.5 -> 1609.6 | Improved |
| use IBCB instead of B-ICB-S for fairness | SBTS top-2 expert · Synthetic fairness (std=0) | CFR: 2679.5 -> 1263.2 | Improved |
Closest papers by meaning
- Inverse Reinforcement Learning by Estimating Expertise of Demonstrators (AAAI 2025 · similarity 0.87)
- Is Behavior Cloning All You Need? Understanding Horizon in Imitation Learning (NeurIPS 2024 · similarity 0.86)
- EvIL: Evolution Strategies for Generalisable Imitation Learning (ICML 2024 · similarity 0.86)
- Reward-free World Models for Online Imitation Learning (ICML 2025 · similarity 0.86)
- Behavior Cloning is Not All You Need: The Optimality of On-Policy Distillation for Noisy Expert Feedback (NeurIPS 2026 · similarity 0.86)
- Recent Advancements in Inverse Reinforcement Learning (AAAI 2024 · similarity 0.86)
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.
