Instruction Lens Score: Your Instruction Contributes a Powerful Object Hallucination Detector for Multimodal Large Language Models

TL;DR

This paper introduces InsLen, a plug-and-play detector for object hallucination in multimodal large language models, leveraging instruction token embeddings to improve detection without extra training.

Contribution

It reveals that instruction token embeddings encode visual info and proposes InsLen, a novel, effective hallucination detection method that outperforms existing approaches.

Findings

InsLen outperforms existing hallucination detection methods.

It works across multiple benchmarks and MLLM architectures.

InsLen does not require auxiliary models or additional training.

Abstract

Multimodal large language models (MLLMs) have achieved remarkable progress, yet the object hallucination remains a critical challenge for reliable deployment. In this paper, we present an in-depth analysis of instruction token embeddings and reveal that they implicitly encode visual information while effectively filtering erroneous information introduced by misleading visual embeddings. Building on this insight, we propose the Instruction Lens Score (InsLen), which combines a Calibrated Local Score with a Context Consistency Score that measures context consistency of the object tokens. The proposed approach serves as a plug-and-play object hallucination detector without relying on auxiliary models or additional training. Extensive experiments across multiple benchmarks and diverse MLLM architectures demonstrate that InsLen consistently outperforms existing hallucination detection methods, highlighting its effectiveness and robustness. The code is available at https://github.com/Fraserlairh/Instruction-Lens-Score.