Research map / Large language models / Alignment and preferences
Reward Constrained Interactive Recommendation with Natural Language Feedback
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 |
|---|---|---|---|
| Replace unconstrained RL with RCR | Interactive recommendation · UT-Zappos50K unseen items | SR@10: 19 % -> 74 % | Helped |
| Replace unconstrained RL with RCR | Interactive recommendation · UT-Zappos50K unseen items | NI: 26.75 ± 1.67 -> 10.91 ± 1.06 | Helped |
| Replace unconstrained RL with RCR | Interactive recommendation · UT-Zappos50K unseen items | NV: 70.02 ± 6.20 -> 11.32 ± 1.98 | Helped |
| Replace hard attribute constraints with learned discriminator constraints | Interactive recommendation · UT-Zappos50K unseen items | SR@10: 52 % -> 74 % | Helped |
| Replace hard attribute constraints with learned discriminator constraints | Interactive recommendation · UT-Zappos50K unseen items | NV: 16.47 ± 2.75 -> 11.32 ± 1.98 | Helped |
| Replace unconstrained RL with RCR | Text generation · Yelp | Test-BLEU-2: 0.807 -> 0.840 | Helped |
| Replace unconstrained RL with RCR | Text generation · Yelp | VR: 40.36% -> 10.49% | Helped |
| Replace unconstrained RL with RCR | Text generation · Yelp | Self-BLEU-2: 0.658 -> 0.683 | Hurt |
| Increase λmax when λmax ≤ 0.05 (ablation) | RCR without discriminator-based item rejection · UT-Zappos50K | NI | Helped |
| Increase λmax when λmax > 0.05 (ablation) | RCR without discriminator-based item rejection · UT-Zappos50K | NI | No clear effect |
| Set λmax = 0.01 (ablation) | RCR without discriminator-based item rejection · UT-Zappos50K | computational cost | Hurt |
Closest papers by meaning
- MTRec: Learning to Align with User Preferences via Mental Reward Models (NeurIPS 2025 · similarity 0.84)
- Semantic Convergence: Harmonizing Recommender Systems via Two-Stage Alignment and Behavioral Semantic Tokenization (AAAI 2025 · similarity 0.82)
- Customizing Language Models with Instance-wise LoRA for Sequential Recommendation (NeurIPS 2024 · similarity 0.82)
- Language Ranker: A Lightweight Ranking framework for LLM Decoding (NeurIPS 2025 · similarity 0.81)
- Think before Recommendation: Autonomous Reasoning-enhanced Recommender (NeurIPS 2025 · similarity 0.81)
- Re2LLM: Reflective Reinforcement Large Language Model for Session-based Recommendation (AAAI 2025 · similarity 0.81)
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.