Research map / Large language models / Retrieval and knowledge
Personalize Before Retrieve: LLM-based Personalized Query Expansion for User-Centric Retrieval
Yingyi Zhang, Pengyue Jia, Derong Xu, Yi Wen, Xianneng Li, Yichao Wang, Wenlin Zhang, Xiaopeng Li, Weinan Gan, Huifeng Guo, Yong Liu, Xiangyu Zhao
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 P-PRF (style-aligned pseudo feedback) (ablation) | PBR · PersonaBench (average) | Recall@5: 0.3005 -> 0.4527 | Improved |
| add P-PRF (ablation) | PBR · PersonaBench Overall, all-MiniLM-L6-v2 | Recall@5: 0.2860 -> 0.4516 | Improved |
| add P-Anchor (graph-based user anchor) (ablation) | PBR · PersonaBench (average) | NDCG@5: 0.3695 -> 0.3819 | Improved |
| add P-Anchor on non-clustered Preference (hard) with multi-qa-MiniLM (ablation) | PBR · PersonaBench Preference (hard), multi-qa-MiniLM | Recall@5: 0.4390 -> 0.4049 | Hurt |
| add pseudo-utterance (roughly) component (ablation) | PBR · PersonaBench | Recall@5 and NDCG@5 | Improved |
| add pseudo-reasoning (logically) component (ablation) | PBR · PersonaBench | Recall@5 and NDCG@5 | Improved |
| use PBR instead of ThinkQE | PBR · PersonaBench (average) | Recall@5: 0.4098 -> 0.4527 | Improved |
| use generic expansion (HyDE) instead of no expansion | HyDE · PersonaBench (average) | Recall@5: 0.4002 -> 0.3440 | Hurt |
| use PBR instead of Base on LongMemEval-s | PBR · LongMemEval-s, bge-base-en-v1.5 | Recall@1: 0.2267 -> 0.2315 | Improved |
| use ThinkQE instead of Base on LongMemEval-s bge | ThinkQE · LongMemEval-s, bge-base-en-v1.5 | Recall@1: 0.2267 -> 0.2124 | Hurt |
| raise semantic threshold theta or neighbor size k2 too high (theta 0.8, k2 10) (ablation) | PBR · PersonaBench | Recall@5 | Hurt |
| moderate propagation (theta 0.65-0.75, k2 5) in P-Anchor (ablation) | PBR · PersonaBench | Recall@5 and NDCG@5 | Improved |
Closest papers by meaning
- SePer: Measure Retrieval Utility Through The Lens Of Semantic Perplexity Reduction (ICLR 2025 · similarity 0.86)
- RAGGED: Towards Informed Design of Scalable and Stable RAG Systems (ICML 2025 · similarity 0.85)
- Improving Consistency in Retrieval Augmented Systems With Group Similarity Rewards (NeurIPS 2026 · similarity 0.84)
- A Theory for Token-Level Harmonization in Retrieval-Augmented Generation (ICLR 2025 · similarity 0.84)
- OneSearch-V2: The Latent Reasoning Enhanced Self-distillation Generative Search Framework (NeurIPS 2026 · similarity 0.84)
- HYDRA: Model Factorization Framework for Black-Box LLM Personalization (NeurIPS 2024 · similarity 0.84)
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.
