Rethinking Refinement: Correcting Generative Bias without Noise Injection
This addresses sample quality issues in generative models for AI researchers, offering an incremental improvement over existing refinement techniques.
The paper tackled the problem of systematic biases in generative models that degrade sample quality, proposing a post-hoc refinement method that achieved a state-of-the-art FID of 1.46 on MNIST from a base FID of 3.95 with only one additional function evaluation.
Generative models, including diffusion and flow-based models, often exhibit systematic biases that degrade sample quality, particularly in high-dimensional settings. We revisit refinement methods and show that effective bias correction can be achieved as a post-hoc procedure, without noise injection or multi-step resampling of the sampling process. We propose a flow-matching-based \textbf{Bi-stage Flow Refinement (BFR)} framework with two refinement strategies operating at different stages: latent space alignment for approximately invertible generators and data space refinement trained with lightweight augmentations. Unlike previous refiners that perturb sampling dynamics, BFR preserves the original ODE trajectory and applies deterministic corrections to generated samples. Experiments on MNIST, CIFAR-10, and FFHQ at 256x256 resolution demonstrate consistent improvements in fidelity and coverage; notably, starting from base samples with FID 3.95, latent space refinement achieves a \textbf{state-of-the-art} FID of \textbf{1.46} on MNIST using only a single additional function evaluation (1-NFE), while maintaining sample diversity.