Recent text-to-image generative models have enabled us to transform our words
into vibrant, captivating imagery. The surge of personalization techniques that
has followed has also allowed us to imagine unique concepts in new scenes.
However, an intriguing question remains: How can we generate a new, imaginary
concept that has never been seen before? In this paper, we present the task of
creative text-to-image generation, where we seek to generate new members of a
broad category (e.g., generating a pet that differs from all existing pets). We
leverage the under-studied Diffusion Prior models and show that the creative
generation problem can be formulated as an optimization process over the output
space of the diffusion prior, resulting in a set of "prior constraints". To
keep our generated concept from converging into existing members, we
incorporate a question-answering model that adaptively adds new constraints to
the optimization problem, encouraging the model to discover increasingly more
unique creations. Finally, we show that our prior constraints can also serve as
a strong mixing mechanism allowing us to create hybrids between generated
concepts, introducing even more flexibility into the creative process.

这篇论文介绍了创造性文本到图像生成的任务，并使用未经研究的扩散先验模型来解决此问题，同时通过一个问题回答模型来逐渐发现越来越独特的创作，最后展示了先验约束不仅作为强大的混合机制，还能引入更多的灵活性。

ConceptLab：使用扩散先验约束的创造性生成

ConceptLab: Creative Generation using Diffusion Prior Constraints

Clickbait is the practice of engineering titles to incentivize readers to
click through to articles. Such titles with sensationalized language reveal as
little information as possible. Occasionally, clickbait will be intentionally
misleading, so natural language processing (NLP) can scan the article and
answer the question posed by the clickbait title, or spoil it. We tackle two
tasks: classifying the clickbait into one of 3 types (Task 1), and spoiling the
clickbait (Task 2). For Task 1, we propose two binary classifiers to determine
the final spoiler type. For Task 2, we experiment with two approaches: using a
question-answering model to identify the span of text of the spoiler, and using
a large language model (LLM) to generate the spoiler. Because the spoiler is
contained in the article, we frame the second task as a question-answering
approach for identifying the starting and ending positions of the spoiler. We
created models for Task 1 that were better than the baselines proposed by the
dataset authors and engineered prompts for Task 2 that did not perform as well
as the baselines proposed by the dataset authors due to the evaluation metric
performing worse when the output text is from a generative model as opposed to
an extractive model.

该研究提出两个二分类器和两种方法，用于分类和揭露标题党，并为第二个任务创建了模型，其中利用问题回答模型来确定标题党的起始和结束位置。