mlsanityData to 14 Oct 2026

Retrieval from Within: An Intrinsic Capability of Attention-Based Models

arXiv 2605.05806 ↗First posted 7 May 2026cs.LGcited by 2

On the map: Retrieval-augmented generation → RAG robustness and adaptive retrieval

Abstract (arXiv)

Retrieval-augmented generation (RAG) typically treats retrieval and generation as separate systems. We ask whether an attention-based encoder-decoder can instead retrieve directly from its own internal representations. We introduce INTRA (INTrinsic Retrieval via Attention), a framework where decoder attention queries score pre-encoded evidence chunks that are then directly reused as context for generation. By construction, INTRA unifies retrieval and generation, eliminating the retriever-generator mismatch typical of RAG pipelines. This design also amortizes context encoding by reusing precomputed encoder states across queries. On question-answering benchmarks, INTRA outperforms strong engineered retrieval pipelines on both evidence recall and end-to-end answer quality. Our results demonstrate that attention-based models already possess a retrieval mechanism that can be elicited, rather than added as an external module.

Key facts. Cited by 2 papers in the map (data through 14 Oct 2026). Placed in RAG robustness and adaptive retrieval, one of 15 directions in Retrieval-augmented generation. Builds on 31 of its 40 references that are in the map. First posted on arXiv 7 May 2026.

Builds on

31 of 40 references.

Closest ancestors · 1

Same direction, newest first

  1. 2023Self-RAG: Learning to Retrieve, Generate, and Critique through Self-Reflection17 Oct

Other references · 30

  1. 2017Attention Is All You Needcited by 60,361
  2. 2020Retrieval-Augmented Generation for Knowledge-Intensive NLP Taskscited by 9,297
  3. 2023Mamba: Linear-Time Sequence Modeling with Selective State Spacescited by 4,718
  4. 2020Dense Passage Retrieval for Open-Domain Question Answeringcited by 3,824
  5. 2018HotpotQA: A Dataset for Diverse, Explainable Multi-hop Question Answeringcited by 3,754
  6. 2017Billion-scale similarity search with GPUscited by 2,497
  7. 2020REALM: Retrieval-Augmented Language Model Pre-Trainingcited by 2,141
  8. 2020Transformers are RNNs: Fast Autoregressive Transformers with Linear Attentioncited by 2,022
  9. 2020Big Bird: Transformers for Longer Sequencescited by 1,684
  10. 2024Transformers are SSMs: Generalized Models and Efficient Algorithms Through Structured State Space Dualitycited by 1,336
  11. 2021Improving language models by retrieving from trillions of tokenscited by 1,293
  12. 2020Constructing A Multi-hop QA Dataset for Comprehensive Evaluation of Reasoning Stepscited by 1,282
18 more
  1. 2025Qwen3 Embedding: Advancing Text Embedding and Reranking Through Foundation Modelscited by 1,261
  2. 2020ColBERT: Efficient and Effective Passage Search via Contextualized Late Interaction over BERTcited by 1,178
  3. 2022Atlas: Few-shot Learning with Retrieval Augmented Language Modelscited by 952
  4. 2021MuSiQue: Multihop Questions via Single-hop Question Compositioncited by 916
  5. 2023C-Pack: Packed Resources For General Chinese Embeddingscited by 578
  6. 2021ColBERTv2: Effective and Efficient Retrieval via Lightweight Late Interactioncited by 476
  7. 2025Search-o1: Agentic Search-Enhanced Large Reasoning Modelscited by 433
  8. 2023Hyena Hierarchy: Towards Larger Convolutional Language Modelscited by 393
  9. 2024Titans: Learning to Memorize at Test Timecited by 328
  10. 2025Agentic Retrieval-Augmented Generation: A Survey on Agentic RAGcited by 249
  11. 2023RA-DIT: Retrieval-Augmented Dual Instruction Tuningcited by 195
  12. 2025NoLiMa: Long-Context Evaluation Beyond Literal Matchingcited by 126
  13. 2024HELMET: How to Evaluate Long-Context Language Models Effectively and Thoroughlycited by 125
  14. 2023InstructRetro: Instruction Tuning post Retrieval-Augmented Pretrainingcited by 53
  15. 2021Adaptive Information Seeking for Open-Domain Question Answeringcited by 29
  16. 2020HopRetriever: Retrieve Hops over Wikipedia to Answer Complex Questionscited by 19
  17. 2023Chain-of-Skills: A Configurable Model for Open-domain Question Answeringcited by 19
  18. 2025T5Gemma 2: Seeing, Reading, and Understanding Longercited by 17

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2026: 2.

2 papers.

  1. 2026MulTaBench: Benchmarking Multimodal Tabular Learning with Text and Imagecited by 2
  2. 2026UNREAL: Unifying Retrieval and Long-Context with a Single Modelcited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2026In-Context Optimization for Retrieval-Augmented Generation: A Gradient-Descent Perspective
  2. 2025Retrieval-Augmented Generation: A Comprehensive Survey of Architectures, Enhancements, and Robustness Frontiers
  3. 2025Scaling Test-Time Inference with Policy-Optimized, Dynamic Retrieval-Augmented Generation via KV Caching and Decoding
  4. 2025Dynamic Context Selection for Retrieval-Augmented Generation: Mitigating Distractors and Positional Bias
  5. 2025Infinite Retrieval: Attention Enhanced LLMs in Long-Context Processing
  6. 2024Quantifying reliance on external information over parametric knowledge during Retrieval Augmented Generation (RAG) using mechanistic analysis
  7. 2023Attention Sorting Combats Recency Bias In Long Context Language Models
  8. 2026Latent Abstraction for Retrieval-Augmented Generation