# Bounding Hallucinations: Merlin-Arthur Protocols for Mutual-Information Bounds in Language Models

arXiv: https://arxiv.org/abs/2512.11614 · HTML: https://mlsanity.com/paper/2512.11614/

## Key facts

- No citing papers in the map yet (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 12 Dec 2025.

## Placed in

- Retrieval-augmented generation / [RAG robustness and adaptive retrieval](https://mlsanity.com/rag/rag-robustness-and-adaptive-retrieval/)

## Builds on (most cited first, 30 of 31)

- [The Llama 3 Herd of Models](https://mlsanity.com/paper/2407.21783/) (2024)
- [LoRA: Low-Rank Adaptation of Large Language Models](https://mlsanity.com/paper/2106.09685/) (2021)
- [Grad-CAM: Visual Explanations from Deep Networks via Gradient-based Localization](https://mlsanity.com/paper/1610.02391/) (2016)
- [SQuAD: 100,000+ Questions for Machine Comprehension of Text](https://mlsanity.com/paper/1606.05250/) (2016)
- [Dense Passage Retrieval for Open-Domain Question Answering](https://mlsanity.com/paper/2004.04906/) (2020)
- [HotpotQA: A Dataset for Diverse, Explainable Multi-hop Question Answering](https://mlsanity.com/paper/1809.09600/) (2018)
- [Axiomatic Attribution for Deep Networks](https://mlsanity.com/paper/1703.01365/) (2017)
- [TriviaQA: A Large Scale Distantly Supervised Challenge Dataset for Reading Comprehension](https://mlsanity.com/paper/1705.03551/) (2017)
- [Know What You Don't Know: Unanswerable Questions for SQuAD](https://mlsanity.com/paper/1806.03822/) (2018)
- [Constructing A Multi-hop QA Dataset for Comprehensive Evaluation of Reasoning Steps](https://mlsanity.com/paper/2011.01060/) (2020)
- [Unsupervised Dense Information Retrieval with Contrastive Learning](https://mlsanity.com/paper/2112.09118/) (2021)
- [MuSiQue: Multihop Questions via Single-hop Question Composition](https://mlsanity.com/paper/2108.00573/) (2021)
- [Generic Attention-model Explainability for Interpreting Bi-Modal and Encoder-Decoder Transformers](https://mlsanity.com/paper/2103.15679/) (2021)
- [Rationalizing Neural Predictions](https://mlsanity.com/paper/1606.04155/) (2016)
- [Understanding Neural Networks through Representation Erasure](https://mlsanity.com/paper/1612.08220/) (2016)
- [Polyjuice: Generating Counterfactuals for Explaining, Evaluating, and Improving Models](https://mlsanity.com/paper/2101.00288/) (2021)
- [Know Your Limits: A Survey of Abstention in Large Language Models](https://mlsanity.com/paper/2407.18418/) (2024)
- [Explaining NLP Models via Minimal Contrastive Editing (MiCE)](https://mlsanity.com/paper/2012.13985/) (2020)
- [ReDeEP: Detecting Hallucination in Retrieval-Augmented Generation via Mechanistic Interpretability](https://mlsanity.com/paper/2410.11414/) (2024)
- [Identifying and Mitigating Spurious Correlations for Improving Robustness in NLP Models](https://mlsanity.com/paper/2110.07736/) (2021)
- [Rejection Improves Reliability: Training LLMs to Refuse Unknown Questions Using RL from Knowledge Feedback](https://mlsanity.com/paper/2403.18349/) (2024)
- [Retrieval-Augmented Generation with Conflicting Evidence](https://mlsanity.com/paper/2504.13079/) (2025)
- [Do LLMs Know When to NOT Answer? Investigating Abstention Abilities of Large Language Models](https://mlsanity.com/paper/2407.16221/) (2024)
- [Uncertainty-Based Abstention in LLMs Improves Safety and Reduces Hallucinations](https://mlsanity.com/paper/2404.10960/) (2024)
- [AtMan: Understanding Transformer Predictions Through Memory Efficient Attention Manipulation](https://mlsanity.com/paper/2301.08110/) (2023)
- [Detecting (Un)answerability in Large Language Models with Linear Directions](https://mlsanity.com/paper/2509.22449/) (2025)
- [Out of Style: RAG's Fragility to Linguistic Variation](https://mlsanity.com/paper/2504.08231/) (2025)
- [Tricking Retrievers with Influential Tokens: An Efficient Black-Box Corpus Poisoning Attack](https://mlsanity.com/paper/2503.21315/) (2025)
- [Interpretability Guarantees with Merlin-Arthur Classifiers](https://mlsanity.com/paper/2206.00759/) (2022)
- [RefusalBench: Generative Evaluation of Selective Refusal in Grounded Language Models](https://mlsanity.com/paper/2510.10390/) (2025)

## Abstract (arXiv)

Retrieval-augmented generation (RAG) relies on retrieved context to guide large language models (LLM), yet treats the retrieval as a heuristic rather than verifiable evidence -- leading to unsupported answers, hallucinations, and reliance on spurious context. We introduce a novel evaluation, data augmentation, and training framework that treats the RAG pipeline as an interactive proof system by adapting the Merlin-Arthur (M/A) protocol: Arthur (the generator LLM) receives context of unknown provenance and Merlin gives helpful evidence, while Morgana injects adversarial, misleading context. We implement them both with an XAI method to self-assess and modify evidence most influential to Arthur. Based on this framework we propose the Explained Information Fraction (EIF) score, that disentangles explanation fidelity from model predictive errors and imperfect benchmarks, and normalizes M/A mutual-information lower bounds to realistic empirical settings. When trained with those contexts, Arthur learns to answer when evidence supports the answer and abstain when evidence is insufficient. Across five QA benchmarks and four LLMs (1B to 32B parameters), M/A reduces incorrect answers under insufficient context by up to 35pp during training and by 18-20pp over vanilla finetuning. Abstention emerges even without any manually annotated unanswerable example or preference pair. We improve EIF-cond by 0.1-0.4 during M/A training and by 0.33-0.38 over vanilla finetuning. Reusing Merlin/Morgana contexts as automatic hard positives and negatives also raises retriever Recall@1 by 2pp. While high accuracy does not guarantee entropy flow from context to answer, our EIF scores -- to our knowledge, the first context-to-answer information bound for RAG generators -- show that autonomous interactive-proof-style supervision enables RAG systems that treat retrieved documents as verifiable evidence.