mlsanityData to 7 Oct 2026

MASH: Modeling Abstention via Selective Help-Seeking

arXiv 2510.01152 ↗First posted 1 Oct 2025cs.CLcited by 3

Not on the map. Why?

Abstract (arXiv)

LLMs cannot reliably recognize their parametric knowledge boundaries and often hallucinate answers to outside-of-boundary questions. In this paper, we introduce MASH (Modeling Abstention via Selective Help-seeking), a training framework that readily extracts abstentions from LLMs. Our key idea is that any external help-seeking by an LLM, i.e. search tool use, can serve as a proxy for abstention if the external help (search) is appropriately penalized while also rewarding answer accuracy. MASH operationalizes this idea using reinforcement learning with a pay-per-search reward. We run experiments on three knowledge-intensive QA datasets. Our results show that MASH substantially improves upon the selective help-seeking performance of prior efficient search approaches; on multi-hop datasets, it improves answer accuracy by 7.6%. Furthermore, MASH demonstrates strong off-the-shelf abstention performance, showcasing behavior competitive with prior abstention methods that additionally require predetermining model knowledge boundaries to construct training data. Overall, we show MASH training effectively aligns search tool use with parametric knowledge, which can be successfully leveraged for making abstention decisions and efficient search tool use

Key facts. Cited by 3 papers in the map (data through 7 Oct 2026). Builds on 35 of its 44 references that are in the map. First posted on arXiv 1 Oct 2025.

Builds on

35 of 44 references.

References · 35

  1. 2023Llama 2: Open Foundation and Fine-Tuned Chat Modelscited by 12,865
  2. 2022ReAct: Synergizing Reasoning and Acting in Language Modelscited by 8,355
  3. 2025Qwen3 Technical Reportcited by 8,195
  4. 2023Direct Preference Optimization: Your Language Model is Secretly a Reward Modelcited by 7,916
  5. 2021Measuring Mathematical Problem Solving With the MATH Datasetcited by 5,203
  6. 2023Mamba: Linear-Time Sequence Modeling with Selective State Spacescited by 4,616
  7. 2023Toolformer: Language Models Can Teach Themselves to Use Toolscited by 4,072
  8. 2020Dense Passage Retrieval for Open-Domain Question Answeringcited by 3,779
  9. 2018HotpotQA: A Dataset for Diverse, Explainable Multi-hop Question Answeringcited by 3,686
  10. 2017TriviaQA: A Large Scale Distantly Supervised Challenge Dataset for Reading Comprehensioncited by 2,895
  11. 2024HybridFlow: A Flexible and Efficient RLHF Frameworkcited by 1,980
  12. 2025Search-R1: Training LLMs to Reason and Leverage Search Engines with Reinforcement Learningcited by 1,371
23 more
  1. 2020Constructing A Multi-hop QA Dataset for Comprehensive Evaluation of Reasoning Stepscited by 1,259
  2. 2022Text Embeddings by Weakly-Supervised Contrastive Pre-trainingcited by 1,216
  3. 2023Instruction-Following Evaluation for Large Language Modelscited by 1,143
  4. 2025Kimi K2: Open Agentic Intelligencecited by 523
  5. 2023Active Retrieval Augmented Generationcited by 506
  6. 2025Cognitive Behaviors that Enable Self-Improving Reasoners, or, Four Habits of Highly Effective STaRscited by 373
  7. 2024INSIDE: LLMs' Internal States Retain the Power of Hallucination Detectioncited by 317
  8. 2024Iterative Reasoning Preference Optimizationcited by 244
  9. 2024Does Fine-Tuning LLMs on New Knowledge Encourage Hallucinations?cited by 227
  10. 2024Don't Hallucinate, Abstain: Identifying LLM Knowledge Gaps via Multi-LLM Collaborationcited by 176
  11. 2025OctoThinker: Mid-training Incentivizes Reinforcement Learning Scalingcited by 110
  12. 2023R-Tuning: Instructing Large Language Models to Say `I Don't Know'cited by 106
  13. 2024SaySelf: Teaching LLMs to Express Confidence with Self-Reflective Rationalescited by 102
  14. 2024Large Language Models Must Be Taught to Know What They Don't Knowcited by 99
  15. 2024HaloScope: Harnessing Unlabeled LLM Generations for Hallucination Detectioncited by 96
  16. 2025Acting Less is Reasoning More! Teaching Model to Act Efficientlycited by 80
  17. 2023Alignment for Honestycited by 72
  18. 2024Rejection Improves Reliability: Training LLMs to Refuse Unknown Questions Using RL from Knowledge Feedbackcited by 64
  19. 2024Can AI Assistants Know What They Don't Know?cited by 47
  20. 2024FLAME: Factuality-Aware Alignment for Large Language Modelscited by 46
  21. 2024Probing-RAG: Self-Probing to Guide Language Models in Selective Document Retrievalcited by 28
  22. 2025TruthRL: Incentivizing Truthful LLMs via Reinforcement Learningcited by 20
  23. 2025Don't lie to your friends: Learning what you know from collaborative self-playcited by 14

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2026: 3.

3 papers.

  1. 2026The Confidence Dichotomy: Analyzing and Mitigating Miscalibration in Tool-Use Agentscited by 14
  2. 2026Capability Self-Assessment in Large Language Modelscited by 1
  3. 2026Don't Start What You Can't Finish: A Counterfactual Audit of Support-State Triage in LLM Agentscited by 0