GOOD: Training-Free Guided Diffusion Sampling for Out-of-Distribution Detection

TL;DR

GOOD introduces a training-free, dual-guided diffusion sampling method that effectively generates diverse out-of-distribution samples, improving detection robustness without relying on perturbing text embeddings.

Contribution

The paper proposes a novel dual-guidance framework for diffusion models that directly guides sampling towards OOD regions using existing classifiers, enhancing OOD sample diversity and detection.

Findings

Improved OOD detection performance with generated samples.

Enhanced diversity and controllability of OOD samples.

Training with GOOD-generated samples boosts detection accuracy.

Abstract

Recent advancements have explored text-to-image diffusion models for synthesizing out-of-distribution (OOD) samples, substantially enhancing the performance of OOD detection. However, existing approaches typically rely on perturbing text-conditioned embeddings, resulting in semantic instability and insufficient shift diversity, which limit generalization to realistic OOD. To address these challenges, we propose GOOD, a novel and flexible framework that directly guides diffusion sampling trajectories towards OOD regions using off-the-shelf in-distribution (ID) classifiers. GOOD incorporates dual-level guidance: (1) Image-level guidance based on the gradient of log partition to reduce input likelihood, drives samples toward low-density regions in pixel space. (2) Feature-level guidance, derived from k-NN distance in the classifier's latent space, promotes sampling in feature-sparse regions. Hence, this dual-guidance design enables more controllable and diverse OOD sample generation. Additionally, we introduce a unified OOD score that adaptively combines image and feature discrepancies, enhancing detection robustness. We perform thorough quantitative and qualitative analyses to evaluate the effectiveness of GOOD, demonstrating that training with samples generated by GOOD can notably enhance OOD detection performance.