Research map / Reinforcement learning / Visual and representation RL
Hierarchical Reinforcement Learning for Integrated Recommendation
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 hierarchical channel selection and item recommendation instead of a flat DDPG agent | DDPG for integrated recommendation · WeChat Top Stories online | ACN: +7.82% -> +10.43% | Helped |
| Add the full HRL-Rec training approach, including supervised click loss | Hierarchical DDPG for integrated recommendation · WeChat Top Stories online | CTR: +6.07% -> +6.34% | Helped |
| Add GRU sequence encoding (ablation) | HRL-Rec · IRec-4B | AUC: 0.7879 -> 0.8097 | Helped |
| Add self-attention feature interaction (ablation) | HRL-Rec · IRec-4B | AUC: 0.8065 -> 0.8097 | Helped |
| Consider future rewards by using a nonzero discount factor (ablation) | HRL-Rec · IRec-4B | AUC: 0.7967 -> 0.8097 | Helped |
| Add similarity loss Ls (ablation) | HRL-Rec · IRec-4B | AUC: 0.8045 -> 0.8097 | Helped |
| Add supervised click loss Lc (ablation) | HRL-Rec · IRec-4B | AUC: 0.8032 -> 0.8097 | Helped |
| Use GRU rather than LR as the state encoder in DQN | DQN recommendation model · WeChat Top Stories online | CTR: +4.17% -> +5.27% | Helped |
| Use Double and Dueling strategies in DQN | DQN recommendation model · WeChat Top Stories online | ACN: +4.77% -> +5.41% | Helped |
| Use DDPG instead of Double-Dueling-DQN | Integrated recommendation model · WeChat Top Stories online | ACN: +5.41% -> +7.82% | Helped |
| Use multi-aspect rewards for diversity and novelty | HRL-Rec · WeChat Top Stories online | AWT: 1.98% | Helped |
| Use hierarchical channel selection and item recommendation to decouple channel and item models | Integrated recommendation model · WeChat Top Stories online, video-channel proportion | Average relative change: 11.7% -> 1.4% | Helped |
Closest papers by meaning
- On the Unexpected Effectiveness of Reinforcement Learning for Sequential Recommendation (ICML 2024 · similarity 0.87)
- Think before Recommendation: Autonomous Reasoning-enhanced Recommender (NeurIPS 2025 · similarity 0.87)
- Contrastive Representation for Interactive Recommendation (AAAI 2025 · similarity 0.86)
- UNEX-RL: Reinforcing Long-Term Rewards in Multi-Stage Recommender Systems with UNidirectional EXecution (AAAI 2024 · similarity 0.86)
- Objective Shaping with Hard Negatives: Windowed Partial AUC Optimization for RL-based LLM Recommenders (NeurIPS 2026 · similarity 0.86)
- ConvRec-R1: Training LLM-based Conversational Recommender Systems with Reinforcement Learning (ICLR 2026 · 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.