A Comprehensive Survey on LLM-Powered Recommender Systems: From Discriminative, Generative to Multi-Modal Paradigms
Bibliographic record
Abstract
Large Language Models (LLMs) have become transformative tools in Natural Language Processing (NLP). They are increasingly being integrated into recommendation systems to address existing limitations such as data sparsity, novelty, cold start, and long-tail challenges. Unlike traditional recommendation techniques that rely on user-item interaction matrices, LLMs provide context-aware reasoning and multi-modal processing capabilities. However, existing research mainly focuses on fine-tuning and prompt engineering strategies without fully exploring hybrid models, retrieval-augmented generation (RAG), graph-enhanced recommendations, and evaluation methodologies. This survey offers a comprehensive and structured examination of LLM-based recommendation systems, categorizing them into discriminative, generative, hybrid, graph-enhanced, and multimodal paradigms. Additionally, we explore adaptive fine-tuning techniques, prompt engineering strategies, and retrieval-augmented generation (RAG) approaches that improve LLM performance in personalized recommendations. We also examine evaluation methodologies, including LLM-as-a-Judge frameworks, benchmark limitations, and fairness considerations. Finally, we present a detailed discussion of open challenges, such as hallucination, scalability, bias, and privacy, highlighting critical research gaps and opportunities for future exploration. This survey aims to guide researchers in navigating the evolving landscape of LLM-driven recommendation systems.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.003 | 0.011 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.002 | 0.001 |
| Bibliometrics | 0.003 | 0.006 |
| Science and technology studies | 0.001 | 0.000 |
| Scholarly communication | 0.002 | 0.004 |
| Open science | 0.002 | 0.002 |
| Research integrity | 0.001 | 0.002 |
| Insufficient payload (model declined to judge) | 0.005 | 0.003 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".