SOURCE-LINKED INTELLIGENCE

Rethinking Diffusion Segmentation: When Does It Rely on Its Noisy State, and Does Diffusion Matter?

arXiv · AI, language, vision and robotics · article · Sep 21, 2026 · UTC

Diffusion models are increasingly adapted from generation to conditional prediction, where a conditioning signal is combined with an evolving noisy representation of the target. In fully supervised segmentation, however, the conditioning image can already support direct target prediction, so endpoint performance alone establishes neither reliance on the added diffusion state nor a deterministic advantage over image-only prediction. For state reliance, we disrupt target-derived state content or correct image-state pairing during retraining of twelve published methods across three datasets, with

Read original source ↗ Open in workspace

recordType

paper

region

Global