Noise optimization during sampling recovers diversity in mode-collapsed diffusion models while preserving output fidelity.
Kingma, Ab- hishek Kumar, Stefano Ermon, and Ben Poole
2 Pith papers cite this work. Polarity classification is still indexing.
2
Pith papers citing it
fields
cs.CV 2verdicts
UNVERDICTED 2representative citing papers
Golden RPG improves compositional text-to-image generation via a region-aware noise predictor with per-region FiLM adapters, injected cross-attention, and adaptive blending, yielding higher cross-region coherence on benchmarks while matching baselines on CLIP metrics.
citing papers explorer
-
It's Never Too Late: Noise Optimization for Collapse Recovery in Trained Diffusion Models
Noise optimization during sampling recovers diversity in mode-collapsed diffusion models while preserving output fidelity.
-
Golden RPG: Confidence-Adaptive Region-Aware Noise for Compositional Text-to-Image Generation
Golden RPG improves compositional text-to-image generation via a region-aware noise predictor with per-region FiLM adapters, injected cross-attention, and adaptive blending, yielding higher cross-region coherence on benchmarks while matching baselines on CLIP metrics.