Prometheus-Vision: Vision-Language Model as a Judge for Fine-Grained Evaluation
This work addresses the problem of fine-grained, transparent evaluation for VLM outputs, providing an accessible tool for researchers and developers, though it is incremental as it adapts existing LM-as-judge methods to the vision-language domain.
The paper tackles the challenge of evaluating long-form responses from Vision-Language Models (VLMs) by proposing Prometheus-Vision, the first open-source VLM evaluator trained on a custom dataset called the Perception Collection, which achieves the highest Pearson correlation with human evaluators and GPT-4V among open-source models.
Assessing long-form responses generated by Vision-Language Models (VLMs) is challenging. It not only requires checking whether the VLM follows the given instruction but also verifying whether the text output is properly grounded on the given image. Inspired by the recent approach of evaluating LMs with LMs, in this work, we propose to evaluate VLMs with VLMs. For this purpose, we present a new feedback dataset called the Perception Collection, encompassing 15K customized score rubrics that users might care about during assessment. Using the Perception Collection, we train Prometheus-Vision, the first open-source VLM evaluator model that can understand the user-defined score criteria during evaluation. Prometheus-Vision shows the highest Pearson correlation with human evaluators and GPT-4V among open-source models, showing its effectiveness for transparent and accessible evaluation of VLMs. We open-source our code, dataset, and model at https://github.com/kaistAI/prometheus-vision