mlsanityData to 7 Oct 2026

Replace, Don't Expand: Mitigating Context Dilution in Multi-Hop RAG via Fixed-Budget Evidence Assembly

arXiv 2512.10787 ↗First posted 11 Dec 2025cs.AI cs.CLcited by 3

On the map: Retrieval-augmented generation → Generation and context integration

Abstract (arXiv)

Retrieval-Augmented Generation (RAG) systems often fail on multi-hop queries when the initial retrieval misses a bridge fact. Prior corrective approaches, such as Self-RAG, CRAG, and Adaptive-k, typically address this by textitadding more context or pruning existing lists. However, simply expanding the context window often leads to context dilution, where distractors crowd out relevant information. We propose SEAL-RAG, a training-free controller that adopts a ``replace, don't expand'' strategy to fight context dilution under a fixed retrieval depth k. SEAL executes a (Search rightarrow Extract rightarrow Assess rightarrow Loop) cycle: it performs on-the-fly, entity-anchored extraction to build a live textitgap specification (missing entities/relations), triggers targeted micro-queries, and uses textitentity-first ranking to actively swap out distractors for gap-closing evidence. We evaluate SEAL-RAG against faithful re-implementations of Basic RAG, CRAG, Self-RAG, and Adaptive-k in a shared environment on HotpotQA and 2WikiMultiHopQA. On HotpotQA (k=3), SEAL improves answer correctness by +3--13 pp and evidence precision by +12--18 pp over Self-RAG. On 2WikiMultiHopQA (k=5), it outperforms Adaptive-k by +8.0 pp in accuracy and maintains 96% evidence precision compared to 22% for CRAG. These gains are statistically significant (p<0.001). By enforcing fixed-k replacement, SEAL yields a predictable cost profile while ensuring the top-k slots are optimized for precision rather than mere breadth. We release our code and data at https://github.com/mosherino/SEAL-RAG.

Key facts. Cited by 3 papers in the map (data through 7 Oct 2026). Placed in Generation and context integration, one of 15 directions in Retrieval-augmented generation. Builds on 21 of its 31 references that are in the map. First posted on arXiv 11 Dec 2025.

Builds on

21 of 31 references.

Closest ancestors · 3

Same direction, newest first

  1. 2023Making Retrieval-Augmented Language Models Robust to Irrelevant Context2 Oct
  2. 2023Active Retrieval Augmented Generation11 May
  3. 2020Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks22 May

Other references · 18

  1. 2023Judging LLM-as-a-Judge with MT-Bench and Chatbot Arenacited by 8,393
  2. 2023Reflexion: Language Agents with Verbal Reinforcement Learningcited by 4,064
  3. 2020Dense Passage Retrieval for Open-Domain Question Answeringcited by 3,779
  4. 2018HotpotQA: A Dataset for Diverse, Explainable Multi-hop Question Answeringcited by 3,686
  5. 2023Lost in the Middle: How Language Models Use Long Contextscited by 3,383
  6. 2023Retrieval-Augmented Generation for Large Language Models: A Surveycited by 2,185
  7. 2022Survey of Hallucination in Natural Language Generationcited by 2,056
  8. 2023Self-RAG: Learning to Retrieve, Generate, and Critique through Self-Reflectioncited by 1,800
  9. 2018Know What You Don't Know: Unanswerable Questions for SQuADcited by 1,538
  10. 2020Constructing A Multi-hop QA Dataset for Comprehensive Evaluation of Reasoning Stepscited by 1,259
  11. 2020ColBERT: Efficient and Effective Passage Search via Contextualized Late Interaction over BERTcited by 1,160
  12. 2022Measuring and Narrowing the Compositionality Gap in Language Modelscited by 1,054
6 more
  1. 2024Adaptive-RAG: Learning to Adapt Retrieval-Augmented Large Language Models through Question Complexitycited by 437
  2. 2023Chain-of-Verification Reduces Hallucination in Large Language Modelscited by 317
  3. 2024Corrective Retrieval Augmented Generationcited by 187
  4. 2019Understanding Dataset Design Choices for Multi-hop Reasoningcited by 50
  5. 2024Are Long-LLMs A Necessity For Long-Context Tasks?cited by 22
  6. 2025Efficient Context Selection for Long-Context QA: No Tuning, No Iteration, Just Adaptive-kcited by 19

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2026: 3.

3 papers.

  1. 2026TASR: Training-Free Adaptive Stopping for Iterative Retrievalcited by 4
  2. 2026No-Worse Context-Aware Decoding: Preventing Neutral Regression in Context-Conditioned Generationcited by 1
  3. 2026AdaGATE: Adaptive Gap-Aware Token-Efficient Evidence Assembly for Multi-Hop Retrieval-Augmented Generationcited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2025Principled Context Engineering for RAG: Statistical Guarantees via Conformal Prediction
  2. 2026REVA: Reusable Evidence View Aggregation for Context-Efficient RAG Serving
  3. 2025Investigating Retrieval-Augmented Generation Systems on Unanswerable, Uncheatable, Realistic, Multi-hop Queries
  4. 2026SCAR: Semantic Continuity-Aware Retrieval for Efficient Context Expansion in RAG
  5. 2026Less can be More: Relieving RAG Bottlenecks via Evidence Frontloading and Pressure-Adaptive Budgeting
  6. 2025Enhancing RAG Efficiency with Adaptive Context Compression
  7. 2024MultiHop-RAG: Benchmarking Retrieval-Augmented Generation for Multi-Hop Queries
  8. 2026Scaling DPPs for RAG: Density Meets Diversity