Research map / Reinforcement learning / Policy gradient and actor-critic
ProRL: Effective Reinforcement Learning for Proactive Recommendation via Rectified Policy Gradient Estimation
Hongru Hou, Tiehua Mei, Denghui Geng, Jinhui Huang, Ao Xu, Hengrui Chen, Jiaqing Liang, Deqing Yang
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 |
|---|---|---|---|
| add Stepwise Reward Centering (ablation) | ProRL · MovieLens-1M | IoI: 1.2373 -> 2.8504 | Improved |
| add Stepwise Reward Centering (ablation) | ProRL · MovieLens-1M | CTR: 0.9731 -> 0.8543 | Hurt |
| add Stepwise Reward Centering (ablation) | ProRL · Steam | IoR: 198.30 -> 340.18 | Improved |
| add Position-Specific Advantage Estimation (ablation) | ProRL · MovieLens-1M | IoI: 2.5556 -> 2.8504 | Improved |
| add Position-Specific Advantage Estimation (ablation) | ProRL · Steam | IoI: 0.7280 -> 1.1188 | Improved |
| add Position-Specific Advantage Estimation (ablation) | ProRL · Amazon-Book | IoR: 1223.78 -> 1383.41 | Improved |
| add IoR reward term (ablation) | ProRL · Amazon-Book | IoR: 125.94 -> 1383.41 | Improved |
| add IoI reward term (ablation) | ProRL · Steam | IoI: 0.2856 -> 1.1188 | Improved |
| add CTR reward term (ablation) | ProRL · MovieLens-1M | CTR: 0.7722 -> 0.8543 | Improved |
| use position-specific advantage instead of REINFORCE (ablation) | policy gradient · MovieLens-1M | IoI: 1.626 -> 2.850 | Improved |
| use position-specific advantage instead of GRPO (ablation) | policy gradient · MovieLens-1M | IoI: 1.483 -> 2.850 | Improved |
| use analytic position-specific baseline instead of learned critic A2C (ablation) | policy gradient · MovieLens-1M | IoR: 527.5 -> 728.2 | Improved |
Later papers that compared with this method
1 later accepted paper reports ProRL as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- BroRL: Scaling Reinforcement Learning via Broadened Exploration (ICML 2026 · better on 2 of 2 shared results)
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- On the Unexpected Effectiveness of Reinforcement Learning for Sequential Recommendation (ICML 2024 · similarity 0.91)
- Objective Shaping with Hard Negatives: Windowed Partial AUC Optimization for RL-based LLM Recommenders (NeurIPS 2026 · similarity 0.89)
- UNEX-RL: Reinforcing Long-Term Rewards in Multi-Stage Recommender Systems with UNidirectional EXecution (AAAI 2024 · similarity 0.88)
- Think before Recommendation: Autonomous Reasoning-enhanced Recommender (NeurIPS 2025 · similarity 0.88)
- MTRec: Learning to Align with User Preferences via Mental Reward Models (NeurIPS 2025 · similarity 0.87)
- Unbiased Recommender Learning from Implicit Feedback via Weakly Supervised Learning (ICML 2025 · similarity 0.87)
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.
