Research map / Reinforcement learning / Policy gradient and actor-critic
ResAct: Reinforcing Long-term Engagement in Sequential Recommendation with Residual Actor
Wanqi Xue, Qingpeng Cai, Ruohan Zhan, Dong Zheng, Peng Jiang, Kun Gai, Bo An
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 a residual actor to improve reconstructed online-policy actions | imitation learning policy · RecL-25m Return Time | NCIS: 0.6404 -> 0.7980 | Improved |
| add a residual actor to improve reconstructed online-policy actions | imitation learning policy · RecL-25m Session Length | NCIS: 0.3186 -> 0.5433 | Improved |
| add a residual actor to improve reconstructed online-policy actions | imitation learning policy · RecL-25m Both | NCIS: 0.5345 -> 0.6675 | Improved |
| use CVAE reconstruction instead of a deterministic behavior reconstructor (ablation) | ResAct · RecL-25m | NCIS | Improved |
| increase the number of online-behavior estimators (ablation) | ResAct · RecL-25m Return Time, Session Length, and Both | NCIS | Improved |
| use behavior-like CVAE candidates instead of adding 20 uniform candidates (ablation) | ResAct · RecL-25m Return Time | NCIS: 0.5501 -> 0.7980 | Improved |
| use behavior-like CVAE candidates instead of adding 20 uniform candidates (ablation) | ResAct · RecL-25m Session Length | NCIS: 0.3489 -> 0.5433 | Improved |
| use behavior-like CVAE candidates instead of adding 20 uniform candidates (ablation) | ResAct · RecL-25m Both | NCIS: 0.4839 -> 0.6675 | Improved |
| add the expressiveness regularizer (ablation) | ResAct · RecL-25m Return Time | NCIS: 0.6610 -> 0.7980 | Improved |
| add the expressiveness regularizer (ablation) | ResAct · RecL-25m Session Length | NCIS: 0.3895 -> 0.5433 | Improved |
| add the expressiveness regularizer (ablation) | ResAct · RecL-25m Both | NCIS: 0.6074 -> 0.6675 | Improved |
| add the conciseness regularizer (ablation) | ResAct · RecL-25m Return Time | NCIS: 0.6944 -> 0.7980 | Improved |
Later papers that compared with this method
2 later accepted papers report Resact as a baseline in their results tables; 2 of them report a better number on most of the results they share with it.
- Multi-timescale Reinforcement Learning by Value Reconstruction (ICML 2026 · better on 4 of 4 shared results)
- Return-Critic: Bridging Goal Discrepancy for Efficient Visual Reinforcement Learning (ICML 2026 · better on 3 of 3 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 0.90)
- UNEX-RL: Reinforcing Long-Term Rewards in Multi-Stage Recommender Systems with UNidirectional EXecution (AAAI 2024 · similarity 0.88)
- MultiScale Contextual Bandits for Long Term Objectives (NeurIPS 2025 · similarity 0.87)
- Re2LLM: Reflective Reinforcement Large Language Model for Session-based Recommendation (AAAI 2025 · 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.
