EVALALIGN: Supervised Fine-Tuning Multimodal LLMs with Human-Aligned Data for Evaluating Text-to-Image Models

TL;DR

EvalAlign is a supervised fine-tuning approach for multimodal large language models that creates accurate, stable, and fine-grained evaluation metrics for text-to-image models, closely aligning with human judgments.

Contribution

The paper introduces EvalAlign, a novel supervised fine-tuning method for MLLMs to produce precise, human-aligned evaluation metrics for text-to-image generation.

Findings

EvalAlign outperforms existing metrics in stability and accuracy.

It closely matches human preferences in model evaluation.

Demonstrated effectiveness across 24 text-to-image models.

Abstract

The recent advancements in text-to-image generative models have been remarkable. Yet, the field suffers from a lack of evaluation metrics that accurately reflect the performance of these models, particularly lacking fine-grained metrics that can guide the optimization of the models. In this paper, we propose EvalAlign, a metric characterized by its accuracy, stability, and fine granularity. Our approach leverages the capabilities of Multimodal Large Language Models (MLLMs) pre-trained on extensive data. We develop evaluation protocols that focus on two key dimensions: image faithfulness and text-image alignment. Each protocol comprises a set of detailed, fine-grained instructions linked to specific scoring options, enabling precise manual scoring of the generated images. We supervised fine-tune (SFT) the MLLM to align with human evaluative judgments, resulting in a robust evaluation model. Our evaluation across 24 text-to-image generation models demonstrate that EvalAlign not only provides superior metric stability but also aligns more closely with human preferences than existing metrics, confirming its effectiveness and utility in model assessment.