method · first seen 4 Oct 2026
test-time alignment
Steering generative model outputs toward human preferences at inference time using a reward model, without retraining.
Mentions per week
PapersCodeDiscussion
Earliest sources
- Best-of-$N$ Guidance for Test-time Diffusion AlignmentPapers · 4 Oct 2026 · Richard Lee Kim, Yeongmin Kim, Gyuwon Sim, Taekyu Kim, Minsang Park, Il-chul Moo…
Latest sources
- Best-of-$N$ Guidance for Test-time Diffusion AlignmentPapers · 4 Oct 2026 · Richard Lee Kim, Yeongmin Kim, Gyuwon Sim, Taekyu Kim, Minsang Park, Il-chul Moo…