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

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 hallucinated responses exhibit smaller deviations from their prompts compared to grounded responses, suggesting that hallucinations often arise from superficial rephrasing rather than substantive reasoning. Leveraging this insight, we propose a model-intrinsic detection method that uses distributional distances as principled hallucination scores, eliminating the need for external knowledge or auxiliary models. To enhance sensitivity, we employ deep learnable kernels that automatically adapt to capture nuanced geometric differences between distributions. Our approach outperforms existing baselines, demonstrating state-of-the-art performance on several benchmarks. The method remains competitive even without kernel training, offering a robust, scalable solution for hallucination detection.

Hallucination Detection in LLMs via Topological Divergence on Attention Graphs

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.

Normalizing self-supervised learning for provably reliable Change Point Detection

Change point detection (CPD) methods aim to identify abrupt shifts in the distribution of input data streams. Accurate estimators for this task are crucial across various real-world scenarios. Yet, traditional unsupervised CPD techniques face significant limitations, often relying on strong assumptions or suffering from low expressive power due to inherent model simplicity. In contrast, representation learning methods overcome these drawbacks by offering flexibility and the ability to capture the full complexity of the data without imposing restrictive assumptions. However, these approaches are still emerging in the CPD field and lack robust theoretical foundations to ensure their reliability. Our work addresses this gap by integrating the expressive power of representation learning with the groundedness of traditional CPD techniques. We adopt spectral normalization (SN) for deep representation learning in CPD tasks and prove that the embeddings after SN are highly informative for CPD. Our method significantly outperforms current state-of-the-art methods during the comprehensive evaluation via three standard CPD datasets.

Universal representations for financial transactional data: embracing local, global, and external contexts

Effective processing of financial transactions is essential for banking data analysis. However, in this domain, most methods focus on specialized solutions to stand-alone problems instead of constructing universal representations suitable for many problems. We present a representation learning framework that addresses diverse business challenges. We also suggest novel generative models that account for data specifics, and a way to integrate external information into a client's representation, leveraging insights from other customers' actions. Finally, we offer a benchmark, describing representation quality globally, concerning the entire transaction history; locally, reflecting the client's current state; and dynamically, capturing representation evolution over time. Our generative approach demonstrates superior performance in local tasks, with an increase in ROC-AUC of up to 14\% for the next MCC prediction task and up to 46\% for downstream tasks from existing contrastive baselines. Incorporating external information improves the scores by an additional 20\%.