Aligning MAGMA by Few-Shot Learning and Finetuning

TL;DR

This paper evaluates and improves the alignment of the MAGMA vision-language model with human values using few-shot learning and finetuning techniques.

Contribution

It introduces a systematic approach to align MAGMA with human values through evaluation, few-shot learning, and adapter-based finetuning.

Findings

Out-of-the-box MAGMA shows baseline alignment.

Few-shot learning improves alignment performance.

Finetuning on aligned examples enhances model behavior.

Abstract

The goal of vision-language modeling is to allow models to tie language understanding with visual inputs. The aim of this paper is to evaluate and align the Visual Language Model (VLM) called Multimodal Augmentation of Generative Models through Adapter-based finetuning (MAGMA) with human values. MAGMA is a VLM that is capable of image captioning and visual question-answering. We will evaluate its alignment in three different scenarios. To begin, we assess MAGMA's out-of-the-box alignment through the checkpoint provided by Hugging Face. Then, we measure if few-shot learning manages to improve the results. Finally, we finetune the model on aligned examples and evaluate its behavior.