Instruction-Following Evaluation of Large Vision-Language Models

TL;DR

This paper investigates why large vision-language models lose instruction-following ability after fine-tuning and shows that explicitly including output format instructions during training can improve their performance.

Contribution

It provides a quantitative analysis of the decline in instruction-following ability post-fine-tuning and demonstrates the effectiveness of including output format instructions during training.

Findings

LVLMs' instruction-following ability declines after fine-tuning.

Including output format instructions improves instruction-following accuracy.

Explicit format instructions during training mitigate performance decline.

Abstract

Following the initial flourishing of large language models (LLMs), there has been a surge in proposed large vision-language models (LVLMs) that integrate LLMs with vision capabilities. However, it has been observed that LVLMs, after tuning to visual instruction using commonly used training datasets, often fail to exhibit the instruction-following ability that was present in the LLM before integration, leading to results in which they do not follow task instructions as expected. This study quantitatively demonstrates that LVLMs' instruction-following ability declines after fine-tuning and analyzes its underlying causes. In particular, we constructed new training datasets highlighting whether the output format is specified. Then, we investigated how explicitly indicating the output format during fine-tuning affects LVLMs' instruction-following ability. Our quantitative evaluation confirmed that LVLMs' instruction-following ability declines after fine-tuning with commonly used datasets. Furthermore, we found that LVLMs trained with datasets, including instructions on output format, tend to follow instructions more accurately than models that do not. These findings suggest that including samples with instructions on output format during (visual) instruction tuning may help mitigate the decline in instruction-following abilities.