mlsanityData to 7 Oct 2026

When Should Multi-Round RAG Stop? Structured Stopping Judgments and Retrieval Reduction in Search-R1

arXiv 2608.13237 ↗First posted 13 Aug 2026cs.IR cs.CLcited by 0

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

Abstract (arXiv)

Multi-round retrieval-augmented generation (RAG) must decide when to stop searching as evidence accumulates. Because the deployed policy is determined by the first STOP on each trajectory, this is a sequential selection problem rather than an independent state-classification task. We adapt S2G-RAG's structured sufficiency-and-gap judgment to a frozen Search-R1 pipeline and train a Qwen3.5-2B judge on 3,009 states from 900 disjoint HotpotQA questions. Search-R1's reasoner, retriever, corpus, prompt, and search budget remain unchanged, while the judge checkpoint and stopping threshold are selected on grouped validation and frozen before confirmatory evaluation. On the confirmatory test set, the resulting policy reduces retrieval calls by 77 (3.70%) relative to Native Search-R1, while Official Exact Match decreases by 0.625 percentage points. Thus, the trained S2G-style structured judge reduces retrieval while broadly preserving answer accuracy. The result does not imply unchanged or improved accuracy, safe stopping, or lower total inference cost.

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Placed in RAG robustness and adaptive retrieval, one of 15 directions in Retrieval-augmented generation. Builds on 10 of its 14 references that are in the map. First posted on arXiv 13 Aug 2026.

Builds on

10 of 14 references.

Closest ancestors · 4

Same direction, newest first

  1. 2026TASR: Training-Free Adaptive Stopping for Iterative Retrieval11 Jun
  2. 2024Adaptive-RAG: Learning to Adapt Retrieval-Augmented Large Language Models through Question Complexity21 Mar
  3. 2024C-RAG: Certified Generation Risks for Retrieval-Augmented Language Models5 Feb
  4. 2023Self-RAG: Learning to Retrieve, Generate, and Critique through Self-Reflection17 Oct

Other references · 6

  1. 2018HotpotQA: A Dataset for Diverse, Explainable Multi-hop Question Answeringcited by 3,686
  2. 2025Search-R1: Training LLMs to Reason and Leverage Search Engines with Reinforcement Learningcited by 1,371
  3. 2022Interleaving Retrieval with Chain-of-Thought Reasoning for Knowledge-Intensive Multi-Step Questionscited by 911
  4. 2017Selective Classification for Deep Neural Networkscited by 799
  5. 2025Knowing You Don't Know: Learning When to Continue Search in Multi-round RAG through Self-Practicingcited by 24
  6. 2026S2G-RAG: Structured Sufficiency and Gap Judging for Iterative Retrieval-Augmented QAcited by 9

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2026When Iterative RAG Beats Ideal Evidence: A Diagnostic Study in Scientific Multi-hop Question Answering
  2. 2025Stop-RAG: Value-Based Retrieval Control for Iterative RAG
  3. 2026Testing Retrieval-Augmented Generation Systems with Chunk Coverage
  4. 2026A Fixed-Budget, Cluster-Aware Standard for LLM-as-a-Judge Evaluation: A Multi-Hop RAG Stress Test
  5. 2026The RAT: A Unified Bayesian Model for RAG Evaluation
  6. 2026BELIEFRAG: Making Adaptive RAG State-Aware under Evolving Evidence
  7. 2026Why RAGs Hallucinate: Penalty-Aware Evaluation of Retrieval-Augmented Generation Systems with Knowledge-Gap Canaries
  8. 2026A Systems-Level Analysis of Sensitivity, Robustness, and Stability in Retrieval-Augmented Generation