Research map / Reinforcement learning / Offline RL
BCORLE($\lambda$): An Offline Reinforcement Learning and Evaluation Framework for Coupons Allocation in E-commerce Market
Yang Zhang, Bo Tang, Qingyu Yang, Dou An, Hongyin Tang, Chenyang Xi, Xueying LI, Feiyu Xiong
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 |
|---|---|---|---|
| replace lambda-update training with lambda-generalization | R-BCQ policy learning · Simulation (stop at AvgCost 7.0, AvgLogins 15) | episodes and training time: 14000 -> 3000 episodes; 2.187h -> 0.439h | Improved |
| add REME OPE instead of DR/IS (simulation) | OPE of R-BCQ · Simulation lambda=0, ls=1000 | AvgLogins error (days): 3.7837 -> 0.0135 | Improved |
| use REME instead of FQE | OPE of R-BCQ · Simulation lambda=0, ls=4000 | AvgLogins error (days): 0.0644 -> 0.0158 | Improved |
| use REME instead of DM | OPE of R-BCQ · Simulation lambda=1, ls=1000 | AvgLogins error (days): 0.0192 -> 0.0100 | Improved |
| use REME instead of IS online | OPE of R-BCQ · Taobao Deals online | AvgLogins error (days): 0.2203 -> 0.0443 | Improved |
| use REME instead of FQE online | OPE of R-BCQ · Taobao Deals online | AvgCost error (Yuan): 0.1515 -> 0.0221 | Improved |
| replace uplift LR+LP with offline RL R-BCQ | coupon allocation · Taobao Deals online week 1 | ROI improvement: 15.47% | Improved |
| replace GBDT+LP with R-BCQ | coupon allocation · Taobao Deals online week 2 | ROI (days/Yuan): 1.3201 -> 1.4773 | Improved |
| swap BCQ for R-BCQ | offline RL algorithm · Taobao Deals online week 1 | ROI (days/Yuan): 1.3950 -> 1.4329 | Improved |
| swap REM for R-BCQ | offline RL algorithm · Taobao Deals online week 2 | AvgLogins (days): 5.9092 -> 5.9871 | Improved |
| swap GBDT+LP for LR+LP (uplift model choice) | uplift baseline · Taobao Deals online week 1 | ROI Imp: -0.31% | No clear effect |
| use model-based MOPO instead of model-free R-BCQ/BCQ/REM | offline RL · Simulation | AvgCost | Hurt |
Closest papers by meaning
- Backdoor Adjustment via Group Adaptation for Debiased Coupon Recommendations (AAAI 2024 · similarity 0.84)
- Online Resource Allocation With General Constraints (NeurIPS 2026 · similarity 0.84)
- Decoupling Learning and Decision-Making: Breaking the $\mathcal{O}(\sqrt{T})$ Barrier in Online Resource Allocation with First-Order Methods (ICML 2024 · similarity 0.84)
- A General Framework for Off-Policy Learning with Partially-Observed Reward (ICLR 2025 · similarity 0.84)
- Non-stochastic Budgeted Online Pricing with Semi-Bandit Feedback (AAAI 2025 · similarity 0.84)
- Near-Optimal Regret-Queue Length Tradeoff in Online Learning for Two-Sided Markets (NeurIPS 2025 · 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.
