홈 › Papers

Hallucination Detection in LLMs via Topological Divergence on Attention Graphs

2025-04-14 · Alexandra Bazarova, Aleksandr Yugay, Andrey Shulga, Alina Ermilova, Andrei Volodichev, Konstantin Polev, Julia Belikova, Rauf Parchiev, Dmitry Simakov, Maxim Savchenko, Andrey Savchenko, Serguei Barannikov, Alexey Zaytsev

Hallucination, i.e., generating factually incorrect content, remains a critical challenge for large language models (LLMs). We introduce TOHA, a TOpology-based HAllucination detector in the RAG setting, which leverages a topological divergence metric to quantify the structural properties of graphs induced by attention matrices. Examining the topological divergence between prompt and response subgraphs reveals consistent patterns: higher divergence values in specific attention heads correlate with hallucinated outputs, independent of the dataset. Extensive experiments, including evaluation on question answering and data-to-text tasks, show that our approach achieves state-of-the-art or competitive results on several benchmarks, two of which were annotated by us and are being publicly released to facilitate further research. Beyond its strong in-domain performance, TOHA maintains remarkable domain transferability across multiple open-source LLMs. Our findings suggest that analyzing the topological structure of attention matrices can serve as an efficient and robust indicator of factual reliability in LLMs.

📄 PDF Abstract BibTeX arXiv:2504.10063

Code (0)

등록된 구현이 없습니다.

Tasks

HallucinationQuestion Answering RAG

Methods 이 논문이 사용한 방법론

Attention 설명 없음

Linear Warmup With Linear Decay Linear Warmup With Linear Decay is a learning rate schedule in which we increase the learning rate linearly for $n$ updates and then linearly decay afterwards.

Dropout Dropout is a regularization technique for neural networks that drops a unit (along with connections) at training time with a specified probability $p$ (a common value is…

Layer Normalization Unlike batch normalization, Layer Normalization directly estimates the normalization statistics from the summed inputs…

BPE Byte Pair Encoding, or BPE, is a subword segmentation algorithm that encodes rare and unknown words as sequences of subword units. The intuition is that various word…

Attention Dropout Attention Dropout is a type of dropout used in attention-based architectures, where elements are randomly dropped out of the…

Softmax The Softmax output function transforms a previous layer's output into a vector of probabilities. It is commonly used for multiclass classification. Given an input vector $x$…

Residual Connection 설명 없음

HalluZig: Hallucination Detection using Zigzag Persistence

2026-01-04 · Shreyas N. Samaga, Gilberto Gonzalez Arroyo, Tamal K. Dey arxiv

The factual reliability of Large Language Models (LLMs) remains a critical barrier to their adoption in high-stakes domains due to their propensity to hallucinate. Current detection methods often rely on surface-level si…

Attention Head Embeddings with Trainable Deep Kernels for Hallucination Detection in LLMs

2025-06-11 · Rodion Oblovatny, Alexandra Bazarova, Alexey Zaytsev

We present a novel approach for detecting hallucinations in large language models (LLMs) by analyzing the probabilistic divergence between prompt and response hidden-state distributions. Counterintuitively, we find that …

Hallucination

The Geometric Price of Discrete Logic: Context-driven Manifold Dynamics of Number Representations

2026-03-24 · Long Zhang, Dai-jun Lin, Wei-neng Chen arxiv

Large language models (LLMs) generalize smoothly across continuous semantic spaces, yet strict logical reasoning demands the formation of discrete decision boundaries. Prevailing theories relying on linear isometric proj…

Logical Reasoning

Cracking the Code of Hallucination in LVLMs with Vision-aware Head Divergence

2024-12-18 · Jinghan He, Kuan Zhu, Haiyun Guo, Junfeng Fang 외

Large vision-language models (LVLMs) have made substantial progress in integrating large language models (LLMs) with visual inputs, enabling advanced multimodal reasoning. Despite their success, a persistent challenge is…

Hallucination Multimodal Reasoning

Attention-guided Self-reflection for Zero-shot Hallucination Detection in Large Language Models

2025-01-17 · Qiang Liu, Xinlong Chen, Yue Ding, Shizhen Xu 외

Hallucination has emerged as a significant barrier to the effective application of Large Language Models (LLMs). In this work, we introduce a novel Attention-Guided SElf-Reflection (AGSER) approach for zero-shot hallucin…