Enhancing Vision-Language Model Reliability with Uncertainty-Guided Dropout Decoding

TL;DR

This paper introduces DROPOUT DECODING, a novel inference-time method that improves vision-language model reliability by quantifying and masking uncertain visual tokens, thereby reducing hallucinations and enhancing output quality.

Contribution

It proposes a new uncertainty-guided token dropout technique applied during inference to mitigate visual misinterpretations in LVLMs, a novel approach compared to traditional training-based methods.

Findings

Significantly reduces object hallucinations in LVLM outputs

Improves reliability and quality of multimodal task performance

Effective across diverse visual benchmarks

Abstract

Large vision-language models (LVLMs) excel at multimodal tasks but are prone to misinterpreting visual inputs, often resulting in hallucinations and unreliable outputs. We present DROPOUT DECODING, a novel inference-time approach that quantifies the uncertainty of visual tokens and selectively masks uncertain tokens to improve decoding. Our method measures the uncertainty of each visual token by projecting it onto the text space and decomposing it into aleatoric and epistemic components. Specifically, we focus on epistemic uncertainty, which captures perception-related errors more effectively. Inspired by dropout regularization, we introduce uncertainty-guided token dropout, which applies the dropout principle to input visual tokens instead of model parameters, and during inference rather than training. By aggregating predictions from an ensemble of masked decoding contexts, we can robustly mitigate errors arising from visual token misinterpretations. Evaluations on benchmarks including CHAIR, THRONE, and MMBench demonstrate that DROPOUT DECODING significantly reduces object hallucinations (OH) and enhances both reliability and quality of LVLM outputs across diverse visual contexts. Code is released at https://github.com/kigb/DropoutDecoding.