Recommender Systems in the Era of Large Language Models (LLMs) — TKDE 2024 Survey

Forum topic · 小凯 · 2026-07-05

Summary

This forum post indexes a survey published in IEEE Transactions on Knowledge and Data Engineering (TKDE, November 2024), titled 'Recommender Systems in the Era of Large Language Models (LLMs)', available via IEEE Xplore behind a subscription paywall. The post provides a structured overview of how large language models are reshaping recommendation systems, covering motivations (efficiency, scalability, and user intent understanding beyond traditional pipeline architectures), method components (representation learning, retrievers, rerankers, planners, generators, and feedback mechanisms), and learning strategies such as supervised fine-tuning, contrastive learning, distillation, and reinforcement learning. It discusses LLM-based paradigms including generative recommendation (Gen-Rec), retrieval-augmented generation (RAG), and agentic search, alongside evaluation practices using datasets like MS MARCO and BEIR with metrics such as nDCG@10 and Recall@k. The post also lists open challenges — evaluation reliability, latency and cost, hallucination and safety, cross-lingual and multimodal generalization — and cross-references related surveys on cross-domain recommendation, retrieval methods, and LLM-powered recommender agents. An engineering checklist addresses data privacy, latency budgets, quality validation, security, and cost control for production deployment.

Recommender Systems in the Era of Large Language Models (LLMs) — TKDE 2024 Survey

Source: IEEE Xplore document 10506571 (subscription required) Venue: IEEE Transactions on Knowledge and Data Engineering (TKDE), November 2024 Category: Recommender Engine Surveys

Key points

Motivation: Traditional recommendation pipelines separate retrieval, ranking, and generation, struggling to meet LLM-era demands for natural-language interaction, multi-hop reasoning, and up-to-date knowledge.
Scope: The survey systematizes work at the intersection of large language models and recommendation, offering a unified, comparable framework for a previously fragmented literature.
Method decomposition: Covers core components including representation learning, retrievers, rerankers, planners, memory modules, generators, and feedback mechanisms.
Learning strategies: Supervised fine-tuning, contrastive learning, distillation, reinforcement learning (including process rewards), and synthetic data bootstrapping.
Emerging paradigms: Discusses interfaces with LLM tool calling, reinforcement learning, and multi-agent collaboration, plus migration paths from research prototypes to industrial systems.
Open problems: Evaluation trustworthiness, latency and cost, hallucination and safety, and cross-lingual / multimodal extension.

Context and related landscape

The post situates the paper within the evolution of neural information retrieval: from BM25 and dense two-tower retrieval, through cross-encoders and late interaction, to generative retrieval and LLM agents. On the recommendation side, it traces the path from matrix factorization and deep CTR models to sequential Transformers, instruction-following LLMs, and generative recommendation (Gen-Rec). Core tensions highlighted include sparse user behavior, huge item catalogs, multi-objective business trade-offs, and the online inference cost and hallucination risks that come with LLMs.

Evaluation practices discussed

Datasets: MS MARCO, BEIR, Natural Questions, domain corpora, public recommendation benchmarks.
Metrics: nDCG@10, MRR, Recall@k, Hit@k, human preference, task success rate, latency, and token cost.
Baselines: BM25, dense retrieval, cross-encoder rerankers, retrieval-free LLMs, commercial search APIs.

Engineering checklist (from the post)

| Concern | Recommendation | |---------|----------------| | Data | PII scrubbing, partitioned indexes, versioned/rollback-capable embeddings | | Latency | Cascades with early stopping, hot-query caching, async reranking | | Quality | Interleaving experiments, human audits, citation verification | | Safety | Source whitelists, adversarial detection, output filtering | | Cost | Small-model routing, distillation, hybrid sparse+dense retrieval |

Cross-referenced related entries

A Comprehensive Review on Harnessing Large Language Models to Overcome… (arXiv 2507.21117)
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 (Gen-Rec), 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)

> Note: Quantitative findings should be verified against the original PDF; this post is based on the abstract and public metadata.

Recommender Systems in the Era of Large Language Models (LLMs) — TKDE 2024 Survey

Summary

Recommender Systems in the Era of Large Language Models (LLMs) — TKDE 2024 Survey

Key points

Context and related landscape

Evaluation practices discussed

Engineering checklist (from the post)

Cross-referenced related entries

Tags