Research map / Large language models / Alignment and preferences
G-UBS: Towards Robust Understanding of Implicit Feedback via Group-Aware User Behavior Simulation
Boyu Chen, Siran Chen, Zhengrong Yue, Kainan Yan, Chenyun Yu, Beibei Kong, Lei Cheng, Chengxiang Zhuo, Zang Li, Yali Wang
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 SFT to Qwen2.5-VL UFM baseline (ablation) | UFM (Qwen2.5-VL-7B) · IF-VR | Person Play Rate: 47.2% -> 48.6% | Improved |
| add RL to Qwen2.5-VL UFM baseline (ablation) | UFM (Qwen2.5-VL-7B) · IF-VR | Reason F1: 30.6% -> 46.0% | Improved |
| add SFT+RL (ablation) | UFM (Qwen2.5-VL-7B) · IF-VR | Judge F1: 35.8% -> 51.4% | Improved |
| add group profile on top of SFT+RL (ablation) | G-UBS UFM · IF-VR | Reason F1: 46.5% -> 55.6% | Improved |
| add reflection step in UGM (ablation) | G-UBS · IF-VR | Person Play Rate: 51.2% -> 52.3% | Improved |
| add visual information in UFM (ablation) | G-UBS · IF-VR | Person Play Rate: 50.9% -> 52.3% | Improved |
| add skip reward to UFM RL (ablation) | G-UBS · IF-VR | Click Rate: 25.2% -> 25.7% | Improved |
| add format reward to UFM RL (ablation) | G-UBS · IF-VR | Person Play Rate: 52.2% -> 52.3% | No clear effect |
| use interest-based grouping instead of demographics-only (ablation) | UGM · IF-VR | Person Play Rate: 52.0% -> 52.3% | Improved |
| add demographics to interest-based grouping (ablation) | UGM · IF-VR | Person Play Rate: 52.3% -> 52.2% | Hurt |
| use TF-IDF instead of BERT-Sim for user clustering (ablation) | UGM · IF-VR | Person Play Rate: 51.8% -> 52.3% | Improved |
| use 20 groups instead of 10 (ablation) | UGM · IF-VR | Person Play Rate: 52.1% -> 52.3% | Improved |
Closest papers by meaning
- MTRec: Learning to Align with User Preferences via Mental Reward Models (NeurIPS 2025 · similarity 0.86)
- LLM-Powered User Simulator for Recommender System (AAAI 2025 · similarity 0.86)
- Exploiting Fine-Grained Skip Behaviors for Micro-Video Recommendation (AAAI 2025 · similarity 0.85)
- VRAgent-R1: Boosting Video Recommendation with MLLM-based Agents via Reinforcement Learning (AAAI 2026 · similarity 0.85)
- UxSID: Semantic-Aware User Interests Modeling for Ultra-Long Sequence (NeurIPS 2026 · similarity 0.85)
- Disentangled Modeling of Preferences and Social Influence for Group Recommendation (AAAI 2025 · similarity 0.85)
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.
