mlsanityData to 14 Oct 2026

Return or Revise? Learning When Revision Helps Retrieval-Augmented QA

arXiv 2609.30087 ↗First posted 24 Sep 2026cs.CL cs.IR cs.LGcited by 1

New in Retrieval-augmented generation; no direction yet. Retrieval-augmented generation

Abstract (arXiv)

We consider the decision of whether to return an existing draft answer or revise it using retrieved evidence, as in answer-revision systems. Draft confidence estimates whether the current answer is correct, but the decision requires estimating the effect of a specified revision. For offline training and evaluation, we grade both the returned draft and its candidate revision under the same correctness judge, which makes repair, harm, and the gap to an oracle observable. We call this paired effect its recoverability, and we train policies to predict it before revision. On 25,870 held-out open-domain questions across three revision setups, a scorer trained on the paired outcome has greater area under the accuracy--revision-rate curve than a matched draft-correctness scorer in all nine Llama setup--seed fits, and gains 0.23--0.68 accuracy points on average at development-selected thresholds, a difference significant across training runs only for dense retrieval. The resulting policy improves on always revising and on average closes more than a third of the oracle gap, although it still applies 38--46% of the harmful revisions. When a draft-free standard-RAG answer is also available, however, choosing between the draft and that answer is stronger by about two points for Llama and four for OLMo, and adding candidate revision as a third option yields no significant gain. Recoverability describes one revision; its value as an available action also depends on the alternatives.

Key facts. Cited by 1 paper in the map (data through 14 Oct 2026). Builds on 44 of its 71 references that are in the map. First posted on arXiv 24 Sep 2026.

Builds on

44 of 71 references.

References · 44

  1. 2024The Llama 3 Herd of Modelscited by 14,715
  2. 2020Retrieval-Augmented Generation for Knowledge-Intensive NLP Taskscited by 9,297
  3. 2023Judging LLM-as-a-Judge with MT-Bench and Chatbot Arenacited by 8,626
  4. 2023Efficient Memory Management for Large Language Model Serving with PagedAttentioncited by 5,944
  5. 2017On Calibration of Modern Neural Networkscited by 4,960
  6. 2020Dense Passage Retrieval for Open-Domain Question Answeringcited by 3,824
  7. 2023Self-Refine: Iterative Refinement with Self-Feedbackcited by 3,442
  8. 2017TriviaQA: A Large Scale Distantly Supervised Challenge Dataset for Reading Comprehensioncited by 2,955
  9. 2017Billion-scale similarity search with GPUscited by 2,497
  10. 2023Self-RAG: Learning to Retrieve, Generate, and Critique through Self-Reflectioncited by 1,839
  11. 2016MS MARCO: A Human Generated MAchine Reading COmprehension Datasetcited by 1,605
  12. 2025gpt-oss-120b & gpt-oss-20b Model Cardcited by 1,386
32 more
  1. 2022When Not to Trust Language Models: Investigating Effectiveness of Parametric and Non-Parametric Memoriescited by 1,128
  2. 2023Large Language Models Cannot Self-Correct Reasoning Yetcited by 1,017
  3. 2023FrugalGPT: How to Use Large Language Models While Reducing Cost and Improving Performancecited by 853
  4. 2017Selective Classification for Deep Neural Networkscited by 852
  5. 2019Latent Retrieval for Weakly Supervised Open Domain Question Answeringcited by 695
  6. 2017Meta-learners for Estimating Heterogeneous Treatment Effects using Machine Learningcited by 478
  7. 2024Adaptive-RAG: Learning to Adapt Retrieval-Augmented Large Language Models through Question Complexitycited by 451
  8. 2024Training Language Models to Self-Correct via Reinforcement Learningcited by 426
  9. 2020Document Ranking with a Pretrained Sequence-to-Sequence Modelcited by 376
  10. 2024Hybrid LLM: Cost-Efficient and Quality-Aware Query Routingcited by 372
  11. 2023Making Retrieval-Augmented Language Models Robust to Irrelevant Contextcited by 369
  12. 2020Consistent Estimators for Learning to Defer to an Expertcited by 271
  13. 2015Recursive Partitioning for Heterogeneous Causal Effectscited by 255
  14. 2020Selective Question Answering under Domain Shiftcited by 238
  15. 2017Predict Responsibly: Improving Fairness and Accuracy by Learning to Defercited by 233
  16. 2023Verify-and-Edit: A Knowledge-Enhanced Chain-of-Thought Frameworkcited by 139
  17. 2024ClashEval: Quantifying the tug-of-war between an LLM's internal prior and external evidencecited by 113
  18. 2023When Does Confidence-Based Cascade Deferral Suffice?cited by 60
  19. 2024Astute RAG: Overcoming Imperfect Retrieval Augmentation and Knowledge Conflicts for Large Language Modelscited by 58
  20. 2025Adaptive Retrieval Without Self-Knowledge? Bringing Uncertainty Back Homecited by 33
  21. 2024Probing-RAG: Self-Probing to Guide Language Models in Selective Document Retrievalcited by 29
  22. 2024Rowen: Adaptive Retrieval-Augmented Generation for Hallucination Mitigation in LLMscited by 27
  23. 2024To Trust or Not to Trust? Enhancing Large Language Models' Situated Faithfulness to External Contextscited by 25
  24. 2025SePer: Measure Retrieval Utility Through The Lens Of Semantic Perplexity Reductioncited by 24
  25. 2024Retrieval Helps or Hurts? A Deeper Dive into the Efficacy of Retrieval Augmentation to Language Modelscited by 19
  26. 2025GainRAG: Preference Alignment in Retrieval-Augmented Generation through Gain Signal Synthesiscited by 12
  27. 2024When to Speak, When to Abstain: Contrastive Decoding with Abstentioncited by 8
  28. 2026Revision or Re-Solving? Decomposing Second-Pass Gains in Multi-LLM Pipelinescited by 7
  29. 2026Is Escalation Worth It? On the Depth of LLM Cascadescited by 4
  30. 2026RouteLMT: Learned Sample Routing for Hybrid LLM Translation Deploymentcited by 2
  31. 2026CounterRefine: Answer-Conditioned Counterevidence Retrieval for Inference-Time Knowledge Repair in Factual Question Answeringcited by 1
  32. 2026Signed Rescue Routing: Harm-Aware Cascades for Efficient LLM Inferencecited by 1

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2026: 1.

1 paper.

  1. 2026When Should an In-Context Learner Expand Its Hypothesis Space?cited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2026When Evidence Changes: Evaluating Memory Repair and Re-reading in Language-Model Agents
  2. 2026Stale-Document Poisoning: When Outdated Retrieval Overrides Correct Model Answers
  3. 2026REVES: REvision and VErification--Augmented Training for Test-Time Scaling
  4. 2023Predicting the Quality of Revisions in Argumentative Writing
  5. 2026What Would Fix This RAG Failure? Auditing Counterfactual Response with Paired Evidence Interventions
  6. 2026CRITIC-R1: Learning Structured Critics for Retrieval-Augmented Generation
  7. 2026Don't Blindly Trust It: How Unreliable Feedback Breaks Tool-Using LLM Agents
  8. 2026Rehearse: Stepping Back from the Confidence Cliff in Self-Improving Autoresearch