Scaling Group Inference for Diverse and High-Quality Generation
This addresses the issue for users in real-world applications who need diverse and high-quality sets of outputs from generative models, representing an incremental advance by optimizing group selection rather than individual samples.
The paper tackles the problem of redundant outputs in generative models when sampling multiple images per prompt, introducing a scalable group inference method that formulates selection as a quadratic integer assignment problem to optimize quality and diversity. Experiments show significant improvements in group diversity and quality across tasks like text-to-image and video generation compared to independent sampling baselines.
Generative models typically sample outputs independently, and recent inference-time guidance and scaling algorithms focus on improving the quality of individual samples. However, in real-world applications, users are often presented with a set of multiple images (e.g., 4-8) for each prompt, where independent sampling tends to lead to redundant results, limiting user choices and hindering idea exploration. In this work, we introduce a scalable group inference method that improves both the diversity and quality of a group of samples. We formulate group inference as a quadratic integer assignment problem: candidate outputs are modeled as graph nodes, and a subset is selected to optimize sample quality (unary term) while maximizing group diversity (binary term). To substantially improve runtime efficiency, we progressively prune the candidate set using intermediate predictions, allowing our method to scale up to large candidate sets. Extensive experiments show that our method significantly improves group diversity and quality compared to independent sampling baselines and recent inference algorithms. Our framework generalizes across a wide range of tasks, including text-to-image, image-to-image, image prompting, and video generation, enabling generative models to treat multiple outputs as cohesive groups rather than independent samples.