# Detecting Hallucinations in Retrieval-Augmented Generation via Semantic-level Internal Reasoning Graph

arXiv: https://arxiv.org/abs/2601.03052 · HTML: https://mlsanity.com/paper/2601.03052/

## Key facts

- Cited by 2 papers in the map (data through 14 Oct 2026).
- Placed in Hallucination detection and mitigation, one of 17 directions in LLM reasoning.
- Builds on 17 of its 45 references that are in the map.
- First posted on arXiv 6 Jan 2026.

## Placed in

- LLM reasoning / [Hallucination detection and mitigation](https://mlsanity.com/lr/hallucination-detection-and-mitigation/)
- Retrieval-augmented generation / [Hallucination detection in RAG](https://mlsanity.com/rag/hallucination-detection-in-rag/)

## Builds on (most cited first, 17 of 17)

- [RoBERTa: A Robustly Optimized BERT Pretraining Approach](https://mlsanity.com/paper/1907.11692/) (2019)
- [Deep Inside Convolutional Networks: Visualising Image Classification Models and Saliency Maps](https://mlsanity.com/paper/1312.6034/) (2013)
- [SmoothGrad: removing noise by adding noise](https://mlsanity.com/paper/1706.03825/) (2017)
- [Discovering Latent Knowledge in Language Models Without Supervision](https://mlsanity.com/paper/2212.03827/) (2022)
- [INSIDE: LLMs' Internal States Retain the Power of Hallucination Detection](https://mlsanity.com/paper/2402.03744/) (2024)
- [Semantic Entropy Probes: Robust and Cheap Hallucination Detection in LLMs](https://mlsanity.com/paper/2406.15927/) (2024)
- [Rethinking with Retrieval: Faithful Large Language Model Inference](https://mlsanity.com/paper/2301.00303/) (2022)
- [Constructing Benchmarks and Interventions for Combating Hallucinations in LLMs](https://mlsanity.com/paper/2404.09971/) (2024)
- [LRP4RAG: Detecting Hallucinations in Retrieval-Augmented Generation via Layer-wise Relevance Propagation](https://mlsanity.com/paper/2408.15533/) (2024)
- [Do LLMs Signal When They're Right? Evidence from Neuron Agreement](https://mlsanity.com/paper/2510.26277/) (2025)
- [Synchronous Faithfulness Monitoring for Trustworthy Retrieval-Augmented Generation](https://mlsanity.com/paper/2406.13692/) (2024)
- [Revisiting LRP: Positional Attribution as the Missing Ingredient for Transformer Explainability](https://mlsanity.com/paper/2506.02138/) (2025)
- [Hallucinate Less by Thinking More: Aspect-Based Causal Abstention for Large Language Models](https://mlsanity.com/paper/2511.17170/) (2025)
- [Distributional Semantics Tracing: A Framework for Explaining Hallucinations in Large Language Models](https://mlsanity.com/paper/2510.06107/) (2025)
- [HalluGuard: Evidence-Grounded Small Reasoning Models to Mitigate Hallucinations in Retrieval-Augmented Generation](https://mlsanity.com/paper/2510.00880/) (2025)
- [Knowing the Facts but Choosing the Shortcut: Understanding How Large Language Models Compare Entities](https://mlsanity.com/paper/2510.16815/) (2025)
- [HalluClean: A Unified Framework to Combat Hallucinations in LLMs](https://mlsanity.com/paper/2511.08916/) (2025)

## Built on by (most cited first, 2 of 2)

- [How Much Do Legal RAG Systems Still Hallucinate?](https://mlsanity.com/paper/2608.14210/) (2026)
- [Locating Answer-Correctness Signals in Frozen Large Language Models](https://mlsanity.com/paper/2609.37700/) (2026)

## Abstract (arXiv)

The Retrieval-augmented generation (RAG) system based on Large language model (LLM) has made significant progress. It can effectively reduce factuality hallucinations, but faithfulness hallucinations still exist. Previous methods for detecting faithfulness hallucinations either neglect to capture the models' internal reasoning processes or handle those features coarsely, making it difficult for discriminators to learn. This paper proposes a semantic-level internal reasoning graph-based method for detecting faithfulness hallucination. Specifically, we first extend the layer-wise relevance propagation algorithm from the token level to the semantic level, constructing an internal reasoning graph based on attribution vectors. This provides a more faithful semantic-level representation of dependency. Furthermore, we design a general framework based on a small pre-trained language model to utilize the dependencies in LLM's reasoning for training and hallucination detection, which can dynamically adjust the pass rate of correct samples through a threshold. Experimental results demonstrate that our method achieves better overall performance compared to state-of-the-art baselines on RAGTruth and Dolly-15k.