Research map / Large language models
Retrieval and knowledge: research map
1,084 papers on Retrieval and knowledge in Large language models, from ICML, NeurIPS, ICLR, CVPR, AAAI, ICCV and ECCV, and the journals TPAMI, IJCV and JMLR (2016–2026), grouped into 2 clusters and 8 approaches. The busiest year so far is 2026.
What it is
Language models store facts in their weights, but that knowledge can be missing, wrong or outdated. Retrieval-augmented generation (RAG) gives the model relevant documents at question time; knowledge editing changes specific facts inside the model; question-answering benchmarks measure both.
Terms you will see on this page:
- RAG
- retrieval-augmented generation: search documents first, then answer using them.
- Knowledge editing
- changing a specific fact a model has learned without retraining it.
- Hallucination
- a fluent but unsupported or false statement.
Start reading with the most cited papers here:
- Wizard of Wikipedia: Knowledge-Powered Conversational Agents (ICLR 2019 · 623 citations)
- Benchmarking Large Language Models in Retrieval-Augmented Generation (AAAI 2024 · 384 citations)
Within Large language models, its share shrank from 11.5% in 2023–24 to 9.6% in 2025–26 (228 → 658 papers at ICML, NeurIPS, ICLR, CVPR and AAAI, the venues with data for all four years).
Explore Retrieval and knowledge in the interactive map
Working on something in this topic? Describe your idea in scime atlas to see which approach it falls under, the closest papers by meaning and how crowded the spot has become.
What changed recently
Approaches whose share of the topic moved most between 2023–24 and 2025–26 (ICML, NeurIPS, ICLR, CVPR and AAAI):
- RAG robustness and evaluation rising: 6% → 23% of the topic (14 → 150 papers)
- Retrieval-augmented reasoning rising: 9% → 14% of the topic (21 → 91 papers)
- Knowledge editing in LLMs rising: 16% → 20% of the topic (36 → 130 papers)
- Explanatory and multiple-choice QA benchmarks falling: 17% → 7% of the topic (38 → 49 papers)
- Knowledge-augmented question answering falling: 22% → 13% of the topic (50 → 84 papers)
Papers from 2025–2026 that are already cited most:
- OmniDocBench: Benchmarking Diverse PDF Document Parsing with Comprehensive Annotations (CVPR 2025 · 29 citations)
Approaches and key papers
Question answering and knowledge editing · 725 papers
Keywords: question, answering, editing
Approaches in this cluster:
- Knowledge graph reasoning with LLMs (180 papers)
Ground LLM reasoning in knowledge graphs through logic-aware and graph-constrained methods. - Knowledge editing in LLMs (170 papers)
Edit or update stored knowledge in language models via parameters and memory modules. - Knowledge-augmented question answering (136 papers)
Inject graph or specialized knowledge into LLM prompts and benchmark question answering. - Explanatory and multiple-choice QA benchmarks (153 papers)
Create datasets and analyses for causal, structured, and multiple-choice question answering. - Domain-specific QA datasets (86 papers)
Build question answering and text-to-SQL resources for medical and scientific domains.
Most cited and most cited since 2024:
- Wizard of Wikipedia: Knowledge-Powered Conversational Agents (ICLR 2019 · 623 citations)
- (Comet-) Atomic 2020: On Symbolic and Neural Commonsense Knowledge Graphs (AAAI 2021 · 293 citations)
- Knowledge Graph Prompting for Multi-Document Question Answering (AAAI 2024 · 155 citations)
- FlexKBQA: A Flexible LLM-Powered Framework for Few-Shot Knowledge Base Question Answering (AAAI 2024 · 77 citations)
Retrieval-augmented generation · 359 papers
Keywords: retrieval augmented, augmented generation, generation rag
Approaches in this cluster:
- Retrieval-augmented reasoning (131 papers)
Improve retrieval-augmented LMs with context compression, search-enhanced reasoning, and generative retrieval. - RAG robustness and evaluation (166 papers)
Benchmark, critique, and harden retrieval-augmented generation against noise and attacks. - Long-context and personalized retrieval (62 papers)
Combine retrieval with long-context handling, calibration, and personalization benchmarks.
Most cited and most cited since 2024:
- Benchmarking Large Language Models in Retrieval-Augmented Generation (AAAI 2024 · 384 citations)
- Retrieval Augmented Language Model Pre-Training (ICML 2020 · 208 citations)
- HippoRAG: Neurobiologically Inspired Long-Term Memory for Large Language Models (NeurIPS 2024 · 105 citations)
- Interpretable Long-Form Legal Question Answering with Retrieval-Augmented Large Language Models (AAAI 2024 · 81 citations)
Related topics in Large language models
- Reasoning and chain of thought (1,574)
- Instruction and fine-tuning (888)
- Agents, code and math (1,748)
- Alignment and preferences (672)
- AI and society (913)
- Language, safety and interpretability (3,002)
