SelfEval: Leveraging the discriminative nature of generative models for evaluation
This provides a reliable and easy-to-use automated evaluation method for diffusion models, addressing limitations of existing metrics that rely on external pretrained models.
The paper tackles the problem of evaluating text alignment in text-to-image generative diffusion models by introducing SelfEval, an automated metric that uses the generative model itself to compute likelihoods for recognition tasks, achieving high agreement with human evaluations across multiple models and benchmarks.
We present an automated way to evaluate the text alignment of text-to-image generative diffusion models using standard image-text recognition datasets. Our method, called SelfEval, uses the generative model to compute the likelihood of real images given text prompts, and the likelihood can be used to perform recognition tasks with the generative model. We evaluate generative models on standard datasets created for multimodal text-image discriminative learning and assess fine-grained aspects of their performance: attribute binding, color recognition, counting, shape recognition, spatial understanding. Existing automated metrics rely on an external pretrained model like CLIP (VLMs) or LLMs, and are sensitive to the exact pretrained model and its limitations. SelfEval sidesteps these issues, and to the best of our knowledge, is the first automated metric to show a high degree of agreement for measuring text-faithfulness with the gold-standard human evaluations across multiple generative models, benchmarks and evaluation metrics. SelfEval also reveals that generative models showcase competitive recognition performance on challenging tasks such as Winoground image-score compared to discriminative models. We hope SelfEval enables easy and reliable automated evaluation for diffusion models.