mlsanityData to 7 Oct 2026

Query Decomposition for RAG: Balancing Exploration-Exploitation

arXiv 2510.18633 ↗First posted 21 Oct 2025cs.AIcited by 5

On the map: Retrieval-augmented generation → RAG pipeline design and evaluation

Abstract (arXiv)

Retrieval-augmented generation (RAG) systems address complex user requests by decomposing them into subqueries, retrieving potentially relevant documents for each, and then aggregating them to generate an answer. Efficiently selecting informative documents requires balancing a key trade-off: (i) retrieving broadly enough to capture all the relevant material, and (ii) limiting retrieval to avoid excessive noise and computational cost. We formulate query decomposition and document retrieval in an exploitation-exploration setting, where retrieving one document at a time builds a belief about the utility of a given sub-query and informs the decision to continue exploiting or exploring an alternative. We experiment with a variety of bandit learning methods and demonstrate their effectiveness in dynamically selecting the most informative sub-queries. Our main finding is that estimating document relevance using rank information and human judgments yields a 35% gain in document-level precision, 15% increase in α-nDCG, and better performance on the downstream task of long-form generation.

Key facts. Cited by 5 papers in the map (data through 7 Oct 2026). Placed in RAG pipeline design and evaluation, one of 15 directions in Retrieval-augmented generation. Builds on 28 of its 40 references that are in the map. First posted on arXiv 21 Oct 2025.

Builds on

28 of 40 references.

Closest ancestors · 3

Same direction, newest first

  1. 2025Auto-ARGUE: LLM-Based Report Generation Evaluation30 Sep
  2. 2025Controlled Retrieval-augmented Context Evaluation for Long-form RAG24 Jun
  3. 2024Long-Context LLMs Meet RAG: Overcoming Challenges for Long Inputs in RAG8 Oct

Other references · 25

  1. 2020Retrieval-Augmented Generation for Knowledge-Intensive NLP Taskscited by 9,160
  2. 2022ReAct: Synergizing Reasoning and Acting in Language Modelscited by 8,355
  3. 2023Direct Preference Optimization: Your Language Model is Secretly a Reward Modelcited by 7,916
  4. 2025Search-R1: Training LLMs to Reason and Leverage Search Engines with Reinforcement Learningcited by 1,371
  5. 2022Decomposed Prompting: A Modular Approach for Solving Complex Taskscited by 554
  6. 2022RLPrompt: Optimizing Discrete Text Prompts with Reinforcement Learningcited by 361
  7. 2019Learning to Retrieve Reasoning Paths over Wikipedia Graph for Question Answeringcited by 147
  8. 2019Answering Complex Open-domain Questions Through Iterative Query Generationcited by 60
  9. 2017Ask the Right Questions: Active Question Reformulation with Reinforcement Learningcited by 44
  10. 2022Transfer Learning Approaches for Building Cross-Language Dense Retrieval Modelscited by 36
  11. 2023NevIR: Negation in Neural Information Retrievalcited by 28
  12. 2024A Survey on Recent Advances in Conversational Data Generationcited by 23
13 more
  1. 2024Translate-Distill: Learning Cross-Language Dense Retrieval by Translation and Distillationcited by 22
  2. 2024Researchy Questions: A Dataset of Multi-Perspective, Decompositional Questions for LLM Web Agentscited by 22
  3. 2024Self-seeding and Multi-intent Self-instructing LLMs for Generating Intent-aware Information-Seeking dialogscited by 18
  4. 2023A Unified Framework for Learned Sparse Retrievalcited by 17
  5. 2025Overview of the TREC 2024 NeuCLIR Trackcited by 13
  6. 2018Efficient Test Collection Construction via Active Learningcited by 7
  7. 2025Reproducing NevIR: Negation in Neural Information Retrievalcited by 7
  8. 2024BoolQuestions: Does Dense Retrieval Understand Boolean Logic in Language?cited by 6
  9. 2025Constructing Set-Compositional and Negated Representations for First-Stage Rankingcited by 5
  10. 2025A Comprehensive Taxonomy of Negation for NLP and Neural Retrieverscited by 4
  11. 2017Active Sampling for Large-scale Information Retrieval Evaluationcited by 3
  12. 2024Efficient data selection employing Semantic Similarity-based Graph Structures for model trainingcited by 2
  13. 2025Beyond Reproducibility: Advancing Zero-shot LLM Reranking Efficiency with Setwise Insertioncited by 2

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2026: 5.

5 papers.

  1. 2026Facet-Level Tracing of Evidence Uncertainty and Hallucination in RAGcited by 2
  2. 2026Learning from AVA: Early Lessons from a Curated and Trustworthy Generative AI for Policy and Development Researchcited by 1
  3. 2026Citation-Closure Retrieval and Per-Rule Attribution for Real-World Regulatory Compliance Question Answeringcited by 1
  4. 2026SubSearch: Intermediate Rewards for Unsupervised Guided Reasoning in Complex Retrievalcited by 0
  5. 2026FinSAgent: Corpus-Aligned Multi-Agent RAG Framework for Evidence-Grounded SEC Filing Question Answeringcited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2024MBA-RAG: a Bandit Approach for Adaptive Retrieval-Augmented Generation through Question Complexity
  2. 2024Stochastic RAG: End-to-End Retrieval-Augmented Generation through Expected Utility Maximization
  3. 2026KohakuRAG: A simple RAG framework with hierarchical document indexing
  4. 2025Fast or Better? Balancing Accuracy and Cost in Retrieval-Augmented Generation with Flexible User Control
  5. 2025From Ranking to Selection: A Simple but Efficient Dynamic Passage Selector for Retrieval Augmented Generation
  6. 2024The Power of Noise: Redefining Retrieval for RAG Systems
  7. 2026Understanding the Behaviors of Environment-aware Information Retrieval
  8. 2026One Size Does Not Fit All! Dynamic Retriever and Generator Selection for RAG