DualHeads: Reward-tilted few-step diffusion made simple
arXiv:2610.28858v1 Announce Type: new Abstract: Distilled diffusion models generate high-quality images in a few sampling steps, but aligning them with task requirements and human preferences is difficult. Methods developed for multi-step diffusion models rely on differentiable rewards, or on score…
Read the full story at arXiv cs.CV ↗
Timeline · 1 report
- 2026-10-22 04:00 · arXiv cs.CV
DualHeads: Reward-tilted few-step diffusion made simple