Survey of Generative Search and Recommendation in the Era of Large Language Models

Forum topic · 小凯 · 2026-07-05

Summary

This arXiv survey (arXiv:2404.16924, April 2024) by Yongqi Li, Xinyu Lin, Wenjie Wang, Fuli Feng, Liang Pang, Wenjie Li and colleagues systematically reviews generative approaches to search and recommendation powered by large language models (LLMs). It organizes the field into four main technical lines: dense retrieval, late-interaction models (e.g., ColBERT), generative information retrieval (directly generating document identifiers or content), and agentic search that models retrieval as sequential decision-making. The survey provides a unified taxonomy covering modeling paradigms, LLM integration patterns (RAG, agents, tool use), optimization objectives (relevance, diversity, freshness), and evaluation methodologies (offline metrics like nDCG/MRR, online A/B tests, and LLM-as-judge). It traces the field's evolution from BERT-based reranking and DPR (2019-2021) through RAG-era fusion (2022-2023) to conversational/agentic search and generative recommender systems (Gen-RecSys). Key open challenges include benchmark reliability, hallucination and safety, latency and cost constraints, cross-lingual and multimodal generalization, and the gap between offline metrics and real user satisfaction. The survey serves as a practical roadmap for researchers and engineers building retrieval-augmented and LLM-driven search and recommendation systems.

A Survey of Generative Search and Recommendation in the Era of Large Language Models

Source: arXiv:2404.16924 (April 2024)
Authors: Yongqi Li, Xinyu Lin, Wenjie Wang, Fuli Feng, Liang Pang, Wenjie Li, et al. (9 authors)
Type: Survey
Category: Recommender Engine Surveys

Key points

Provides a unified perspective on generative search and recommendation (generative IR and Gen-RecSys) in the LLM era, consolidating previously scattered work into a comparable framework.
Decomposes system components: representation learning, retrievers, rerankers, planners, generators, and feedback mechanisms.
Contrasts four major technical lines:
Dense retrieval: high recall, low latency; suitable for first-stage retrieval.
Late interaction (e.g., ColBERT): higher accuracy but larger index footprint.
Generative IR: directly "generates" documents via tokens or docids, simplifying the pipeline but facing index-update difficulties.
Agentic search: models search as sequential decision-making, supporting multi-hop reasoning and self-reflection.

Taxonomy

| Dimension | Subcategories | Representative ideas | Strengths | Limitations | |---|---|---|---|---| | Modeling paradigm | Discriminative / generative retrieval | Bi-encoders, cross-encoders, DSI, GPT indexing | Mature, scalable | Semantic drift, update cost | | LLM integration | RAG / Agent / Tool-use | Retrieval augmentation, search agents, API calls | Flexible, interpretable | Latency, error propagation | | Optimization objectives | Relevance / diversity / freshness | Multi-objective LTR, RLHF, online learning | Business-aligned | Scarce annotations | | Evaluation | Offline / online / human | nDCG, MRR, LLM-as-judge, A/B tests | Comparable | Deviates from real satisfaction |

Evolution timeline

2019–2021: BERT reranking and DPR establish neural retrieval foundations.
2022–2023: RAG and FreshLLM drive retrieval-generation fusion.
2024 onward: Conversational/agentic search and Gen-RecSys expand rapidly.
2025–2026 outlook: RL-trained search agents, Deep Research, and GraphRAG become growth frontiers.

Evaluation paradigms covered

Datasets: MS MARCO, BEIR, Natural Questions, recommendation benchmarks, domain corpora.
Metrics: nDCG@10, MRR, Recall@k, Hit@k, human preference, task success rate, latency and token cost.
Baselines: BM25, dense retrieval, cross-encoder reranking, retrieval-free LLMs, commercial search APIs.
Ablations: contribution of retrieval steps, reranking depth, and training data scale.

Takeaways for search / recommendation practitioners

1. Architecture: Cascaded retrieve–rerank–generate remains mainstream, but agentic paradigms make retrieval strategy itself a learnable object. 2. Data: High-quality instruction data and click/session logs matter; synthetic data risks 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 evaluation. 4. Productization: Latency, cost, explainability, and safety are hard constraints for industrial deployment—do not optimize academic benchmarks alone.

Open problems

Lack of unified benchmarks; irreproducible private data.
LLM evaluation bias.
Safety and cost constraints for agentic systems.
Future directions: fine-grained process supervision, joint retrieval–reasoning training, multimodal and cross-lingual consistency, deeper integration with knowledge graphs, and causal/fairness constraints in recommendation.

Glossary

| Term | Meaning | |---|---| | RAG | Retrieval-Augmented Generation | | LTR | Learning to Rank | | nDCG | Normalized Discounted Cumulative Gain | | Agentic Search | Search modeled as sequential decisions and tool calls | | Gen-IR | Generative Information Retrieval |

Related entries

A Comprehensive Survey on Retrieval Methods in Recommender Systems (arXiv:2407.21022)
A Survey on LLM-powered Agents for Recommender Systems (arXiv:2502.10050)
A Review of Modern Recommender Systems Using Generative Models (Gen-RecSys, ACM)
A Comprehensive Survey on Cross-Domain Recommendation (arXiv:2503.14110)
A Comprehensive Review on Harnessing Large Language Models (arXiv:2507.21117)