A Survey of Model Architectures in Information Retrieval
Source Metadata
- Authors: Zhichao Xu, Fengran Mo, Zhiqi Huang, Crystina Zhang, Puxuan Yu, Bei Wang, et al. (8 authors)
- Published: 2025-02-20
- Link: https://arxiv.org/abs/2502.14822
- Type: Survey
- Category: Search Surveys
- Provides a unified perspective that brings fragmented IR work into a comparable framework.
- Decomposes system components (representation learners, retrievers, rerankers, planners, generators, feedback mechanisms) for engineering clarity.
- Offers reproducible benchmarks, datasets, and classification schemes that lower the entry cost for new researchers.
- Discusses interfaces with LLM tool use, reinforcement learning, and multi-agent collaboration, plus paths from research prototypes to industrial systems.
- Lists open problems including evaluation trustworthiness, latency and cost, hallucination and safety, and cross-lingual and multimodal extension.
- Dense retrieval: high recall, low latency; suited to first-stage retrieval.
- Late interaction (e.g., ColBERT): higher precision at the cost of larger indexes.
- Generative IR: emits tokens or docids to "generate" documents, simplifying cascades.
- Agentic search: casts search as sequential decision-making, supporting multi-hop reasoning and self-reflection.
- 2019–2021: BERT reranking and Dense Passage Retrieval (DPR) lay the foundation for neural retrieval.
- 2022–2023: Retrieval-Augmented Generation (RAG) and FreshLLM drive retrieval–generation fusion.
- 2024 onward: Conversational / agentic search and generative recommender systems (Gen-RecSys) expand rapidly.
- 2025–2026: RL-trained search agents, Deep Research systems, and GraphRAG become emerging growth areas.
- Datasets: MS MARCO, BEIR, Natural Questions, domain corpora, public recommendation sets.
- Metrics: nDCG@10, MRR, Recall@k, Hit@k, human preference, task success rate, latency and token cost.
- Baselines: BM25, dense retrievers, cross-encoder rerankers, retrieval-free LLMs, commercial search APIs.
- Ablations: Effects of retrieval steps, reranking depth, and training-data scale on final quality.
One-Sentence Summary
This work systematically reviews the evolution of model architectures in information retrieval (IR), tracing the shift from term-based methods through neural retrievers to LLM-driven and agentic systems.
Background and Motivation
Large-scale search, recommendation, and personalization systems face persistent challenges in efficiency, scalability, and user-intent understanding. Traditional pipelines treat retrieval, ranking, and generation as separate stages, which is poorly matched to LLM-era demands for natural-language interaction, multi-hop reasoning, and real-time knowledge. The survey addresses this gap by mapping how backbone models and end-to-end architectures have co-evolved.
From the problem-definition perspective, the paper covers open-domain information access, enterprise search, semantic understanding in recommendation, and end-to-end architectures that couple external knowledge sources with generative models. The English abstract notes:
> The period from 2019 to the present marks one of the most significant paradigm shifts in information retrieval (IR) and natural language processing (NLP), culminating in the emergence of powerful large language models (LLMs) from 2022 onward. Methods based on pretrained encoder-only architectures (e.g., BERT) as well as decoder-only generative LLMs have outperformed many earlier approaches, demonstrating particularly strong performance in zero-shot scenarios and complex reasoning tasks. This survey examines the evolution of model architectures in IR, with a focus on two key aspects: backbone models for feature extraction and end-to-end system architectures for relevance estimation.
Core Contributions
Taxonomy
| Dimension | Subcategory | Representative Approaches | Strengths | Limitations | |-----------|-------------|---------------------------|-----------|-------------| | Modeling paradigm | Discriminative / Generative retrieval | Dual encoders, cross-encoders, DSI, GPT-indexed | Mature, scalable | Semantic drift, update cost | | LLM integration | RAG / Agent / Tool-use | Retrieval augmentation, search agents, API calls | Flexible, interpretable | Latency, error propagation | | Optimization objective | Relevance / Diversity / Freshness | Multi-objective LTR, RLHF, online learning | Business-aligned | Annotation scarcity | | Evaluation | Offline / Online / Human | nDCG, MRR, LLM-as-judge, A/B | Comparable | Gap with real satisfaction |
Representative Lineages
The survey treats four main lines in parallel:
Timeline of Research Evolution
Experimental and Evaluation Paradigms
For a survey, the covered evaluation design typically includes:
Key Conclusions and Insights
1. Architecture: Cascaded retrieve-then-rerank-then-generate is still mainstream, but agentic paradigms increasingly treat "how many retrieval steps and which strategy" as a learnable object. 2. Data: High-quality instruction data and click/session logs are equally critical; synthetic data must guard against distribution shift and knowledge leakage. 3. Evaluation: Offline metrics diverge from online satisfaction; LLM-as-judge should be cross-validated with human evaluation. 4. Production: Latency, cost, interpretability, and safety are hard industrial constraints and cannot be ignored in favor of academic benchmarks.
Research Gaps and Future Directions
Authors highlight the absence of unified benchmarks, reproducibility issues with private data, LLM evaluation bias, and safety/cost constraints in agentic systems. Future work includes finer-grained process supervision, joint retrieval–reasoning training, enterprise metadata governance, and multimodal and cross-lingual consistency.
Engineering Deployment Checklist
| Item | Question | Recommendation | |------|----------|----------------| | Data | Does training/indexing include PII? How are versions handled? | Partitioned indexes, PII redaction, versioned embeddings with rollback | | Latency | p99 budget? How many retrieval steps? | Cascade with early stopping, hot-query caching, async reranking | | Quality | Does offline improvement translate to online CTR/satisfaction? | Interleaving experiments, sample audits, citation checks | | Safety | Does open retrieval introduce poisoning or bias? | Source whitelists, adversarial detection, output filtering | | Cost | Token and GPU usage per query? | Model routing, distillation, sparse+dense hybrid |
Glossary
| Term | Meaning | |------|---------| | IR | Information Retrieval | | RAG | Retrieval-Augmented Generation | | LTR | Learning to Rank | | nDCG | Normalized Discounted Cumulative Gain | | Agentic Search | Paradigm that models search as sequential decision-making with tool use | | Gen-IR | Generative Information Retrieval |
Original Abstract
> The period from 2019 to the present marks one of the most significant paradigm shifts in information retrieval (IR) and natural language processing (NLP), culminating in the emergence of powerful large language models (LLMs) from 2022 onward. Methods based on pretrained encoder-only architectures (e.g., BERT) as well as decoder-only generative LLMs have outperformed many earlier approaches, demonstrating particularly strong performance in zero-shot scenarios and complex reasoning tasks. This survey examines the evolution of model architectures in IR, with a focus on two key aspects: backbone models for feature extraction and end-to-end system architectures for relevance estimation. To maintain analytical clarity, we deliberately separate architectural design from training methodologies, enabling a focused examination of structural innovations in IR systems. We trace the progression from traditional term-based retrieval models to modern neural approaches, highlighting the transformative impact of transformer-based architectures and subsequent LLM developments. The survey concludes with a forward-looking discussion of open challenges and emerging research directions, including architectural optimization for efficiency and scalability, robust handling of multimodal and multilingual data, and adaptation to novel application domains such as autonomous search agents, which may represent the next paradigm in IR.