Max-pooling Network Revisited: Analyzing the Role of Semantic Probability in Multiple Instance Learning for Hallucination Detection
It addresses the computational overhead of existing hallucination detection methods for LLMs, offering a simpler and faster alternative.
The paper proposes a lightweight hallucination detection method for LLMs that uses max-pooling and an MLP to aggregate token-level features, achieving efficiency gains (no semantic consistency computations) while maintaining competitive performance with state-of-the-art methods like HaMI.
Hallucination detection has become increasingly important for improving the reliability of large language models (LLMs). Recently, hybrid approaches such as HaMI, which combine semantic consistency with internal model states via Multiple Instance Learning (MIL), have achieved state-of-the-art performance. However, these methods incur substantial computational overhead due to repeated sampling and costly semantic similarity computations. In this work, we first provide a theoretical analysis of HaMI in terms of decision margins, revealing that scaling internal states with semantic consistency leads to an enlarged decision margin. Motivated by this insight, we revisit classical sentence classification models from a margin enlargement perspective, aggregating token-level features via max pooling and directly estimating sentence scores using a lightweight MLP. Without requiring semantic consistency computations, our approach achieves substantial efficiency improvements while maintaining competitive performance with state-of-the-art baselines through adaptive aggregation of internal feature representations.