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

2026-05-12

AI Summary

This work addresses the reliability challenges posed by object hallucination in multimodal large language models (MLLMs). It reveals for the first time that instruction token embeddings implicitly encode visual information capable of filtering out misleading signals. Building on this insight, the authors propose a plug-and-play hallucination detection method that requires neither additional training nor auxiliary models. The approach integrates calibrated local scores with contextual consistency scores and is applicable across diverse MLLM architectures. Extensive evaluations demonstrate that the method significantly outperforms existing techniques on multiple benchmarks, exhibiting exceptional effectiveness and robustness.

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.

Problem

• object hallucination
• multimodal large language models
• reliable deployment
• visual grounding
• hallucination detection

Innovation

• Instruction Lens Score
• object hallucination detection
• multimodal large language models
• instruction token embeddings
• context consistency