NoiseBoost: Alleviating Hallucination with Noise Perturbation for Multimodal Large Language Models

TL;DR

NoiseBoost is a simple yet effective method that reduces hallucinations in multimodal large language models by using noise perturbations to balance attention between visual and linguistic information, improving accuracy and enabling semi-supervised learning.

Contribution

The paper introduces NoiseBoost, a novel noise perturbation technique that alleviates hallucinations in MLLMs and enables semi-supervised learning, with consistent performance improvements across training strategies.

Findings

Improves dense caption accuracy by 8.1% with human evaluation.

Enables semi-supervised learning using unlabeled data.

Achieves comparable results with half the data.

Abstract

Multimodal large language models (MLLMs) contribute a powerful mechanism to understanding visual information building on large language models. However, MLLMs are notorious for suffering from hallucinations, especially when generating lengthy, detailed descriptions for images. Our analysis reveals that hallucinations stem from the inherent summarization mechanism of large language models, leading to excessive dependence on linguistic tokens while neglecting vision information. In this paper, we propose NoiseBoost, a broadly applicable and simple method for alleviating hallucinations for MLLMs through the integration of noise feature perturbations. Noise perturbation acts as a regularizer, facilitating a balanced distribution of attention weights among visual and linguistic tokens. Despite its simplicity, NoiseBoost consistently enhances the performance of MLLMs across common training strategies, including supervised fine-tuning and reinforcement learning. Further, NoiseBoost pioneerly enables semi-supervised learning for MLLMs, unleashing the power of unlabeled data. Comprehensive experiments demonstrate that NoiseBoost improves dense caption accuracy by 8.1% with human evaluation and achieves comparable results with 50% of the data by mining unlabeled data. Code and models are available at https://kaiwu5.github.io/noiseboost.