Scaling Group Inference for Diverse and High-Quality Generation

2025-08-21

AI Summary

Independent sampling in generative models yields redundant multi-sample outputs, severely hindering user selection and creative exploration. To address this, we propose a scalable group inference framework that formulates diverse sample generation as a quadratic integer assignment problem, where candidate samples are represented as graph nodes and jointly optimized for both global quality and intra-group diversity. We introduce a novel intermediate-prediction-guided progressive pruning strategy, enabling efficient identification of high-quality, high-diversity solutions from large candidate pools. Our method is unified across text-to-image, image-to-image, image-conditioned, and video generation tasks. It preserves single-sample fidelity while significantly enhancing group-level diversity and practical utility—outperforming both independent sampling and existing group inference algorithms across all evaluated benchmarks.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Natural Language Processing: Generation
• Search and Optimization: Sampling/Simulation-based Search

Application Category

• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

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.

Problem

• Optimizing group diversity and quality in generative models
• Reducing redundancy in multiple output samples per prompt
• Solving quadratic integer assignment for scalable group inference

Innovation

• Scalable group inference method
• Quadratic integer assignment problem
• Progressively prune candidate set