Inference-time Scaling of Diffusion Models through Classical Search
This work addresses the problem of flexible and efficient inference-time adaptation in diffusion models for AI applications, representing a novel method rather than an incremental improvement.
The paper tackled the challenge of inference-time control in diffusion models by using classical search algorithms to adapt generated outputs for test-time objectives, resulting in significant gains in performance and efficiency across tasks like planning, offline reinforcement learning, and image generation.
Classical search algorithms have long underpinned modern artificial intelligence. In this work, we tackle the challenge of inference-time control in diffusion models -- adapting generated outputs to meet diverse test-time objectives -- using principles from classical search. We propose a general framework that orchestrates local and global search to efficiently navigate the generative space. It employs a theoretically grounded local search via annealed Langevin MCMC and performs compute-efficient global exploration using breadth-first and depth-first tree search. We evaluate our approach on a range of challenging domains, including planning, offline reinforcement learning, and image generation. Across all tasks, we observe significant gains in both performance and efficiency. These results show that classical search provides a principled and practical foundation for inference-time scaling in diffusion models. Project page at https://diffusion-inference-scaling.github.io/.