A Comprehensive Review on Harnessing Large Language Models to Overcome Recommender System Challenges
Source: arXiv:2507.21117 — published 2025-07-17. Authors: Rahul Raja, Anshaj Vats, Arpita Vats, Anirban Majumder. Type: Survey.
TL;DR
This survey systematically reviews how Large Language Models (LLMs) can be leveraged to tackle the core bottlenecks of conventional recommender systems, arguing that LLMs are foundational enablers—not mere add-ons—for building adaptive, semantically rich, and user-centric recommenders.
Original Abstract (excerpts)
> Recommender systems have traditionally followed modular architectures comprising candidate generation, multi-stage ranking, and re-ranking, each trained separately with supervised objectives and hand-engineered features. While effective in many domains, such systems face persistent challenges including sparse and noisy interaction data, cold-start problems, limited personalization depth, and inadequate semantic understanding of user and item content. The recent emergence of Large Language Models (LLMs) offers a new paradigm for addressing these limitations through unified, language-native mechanisms that can generalize across tasks, domains, and modalities...
Key points
- Problem addressed: Traditional modular recommendation pipelines (candidate generation → ranking → re-ranking) struggle with sparse/noisy interaction data, cold-start, shallow personalization, and poor semantic understanding of content.
- LLM techniques reviewed:
- Prompt-driven candidate retrieval
- Language-native ranking
- Retrieval-augmented generation (RAG)
- Conversational recommendation
- Strengths of LLM integration: enhanced personalization, semantic alignment, and interpretability without extensive task-specific supervision; zero- and few-shot reasoning enables cold-start and long-tail operation via external knowledge and contextual cues.
- Contribution: A structured framework (taxonomy) of the design space of LLM-enhanced recommenders, plus analysis of trade-offs between accuracy, scalability, and real-time performance.
- Dense retrieval: high recall, low latency, suited to first-stage retrieval;
- Late interaction (e.g., ColBERT): higher accuracy but larger indexes;
- Generative IR: directly "generates" documents via tokens or doc IDs, simplifying cascades;
- Agentic search: models search as sequential decision-making, supporting multi-hop reasoning and self-reflection.
- 2019–2021: BERT re-ranking and DPR establish neural retrieval foundations;
- 2022–2023: RAG and FreshLLM drive retrieval–generation fusion;
- 2024 onward: conversational/agentic search and Gen-RecSys proliferate;
- 2025–2026: RL-trained search agents, Deep Research, and GraphRAG become growth frontiers.
- Lack of unified benchmarks; private data hinders reproducibility
- LLM evaluation bias
- Safety and cost constraints of agentic systems
- Finer-grained process supervision, joint retrieval–reasoning training, multimodal and cross-lingual consistency, causal and fairness constraints
- A Comprehensive Survey on Cross-Domain Recommendation (arXiv:2503.14110)
- A Comprehensive Survey on Retrieval Methods in Recommender Systems (arXiv:2407.21022)
- A Review of Modern Recommender Systems Using Generative Models (DOI: 10.1145/3637528.3671474)
- A Survey of Generative Search and Recommendation in the Era of Large Language Models (arXiv:2404.16924)
- A Survey on LLM-powered Agents for Recommender Systems (arXiv:2502.10050)
Taxonomy and Technical Landscape
| Dimension | Sub-categories | Representative ideas | Pros | Limitations | |---|---|---|---|---| | Modeling paradigm | Discriminative vs. generative retrieval | Two-tower, cross-encoders, DSI, generative doc IDs | Mature, scalable | Semantic drift, update cost | | LLM integration | RAG / Agent / Tool-use | Retrieval-augmented, search agents, API calls | Flexible, interpretable | Latency, error propagation | | Optimization targets | Relevance / diversity / freshness | Multi-objective LTR, RLHF, online learning | Business-aligned | Scarce labels | | Evaluation | Offline / Online / Human | nDCG, MRR, LLM-as-judge, A/B tests | Comparable | Gap with real user satisfaction |
Four technical lines are commonly contrasted:
Research Timeline
Conclusions and Insights
1. Architecture: Cascade retrieval + re-rank + generation remains mainstream, but agentic paradigms make retrieval strategy itself learnable. 2. Data: High-quality instruction data and click/session logs are both critical; synthetic data must avoid knowledge leakage and distribution shift. 3. Evaluation: The gap between offline metrics and online satisfaction is widening; LLM-as-judge needs cross-validation with human assessment. 4. Deployment: Latency, cost, interpretability, and safety are hard industrial constraints—do not optimize academic benchmarks alone.