Do we Really Need Visual Instructions? Towards Visual Instruction-Free Fine-tuning for Large Vision-Language Models

TL;DR

This paper introduces ViFT, a novel framework for fine-tuning large vision-language models without visual instructions, using only text instructions and image captions, achieving state-of-the-art results efficiently.

Contribution

ViFT enables visual instruction-free fine-tuning of LVLMs, reducing data requirements and inheriting task-solving capabilities from backbone LLMs.

Findings

Achieves state-of-the-art performance on visual reasoning benchmarks.

Requires less training data compared to traditional methods.

Effectively combines text and image representations during inference.

Abstract

Visual instruction tuning has become the predominant technology in eliciting the multimodal task-solving capabilities of large vision-language models (LVLMs). Despite the success, as visual instructions require images as the input, it would leave the gap in inheriting the task-solving capabilities from the backbone LLMs, and make it costly to collect a large-scale dataset. To address it, we propose ViFT, a visual instruction-free fine-tuning framework for LVLMs. In ViFT, we only require the text-only instructions and image caption data during training, to separately learn the task-solving and visual perception abilities. During inference, we extract and combine the representations of the text and image inputs, for fusing the two abilities to fulfill multimodal tasks. Experimental results demonstrate that ViFT can achieve state-of-the-art performance on several visual reasoning and visual instruction following benchmarks, with rather less training data. Our code and data will be publicly released.