Diffusion models pull early samples toward dataset center even where no data exists
alec_helbling · x · 2026-09-08
The author points out a curious property of diffusion models: when a sample is mostly noise, the model's best guess is close to the mean, so early in sampling, samples can be pulled toward the center of the dataset—even where no data exists—before rebounding toward the data manifold.
More from Research
- DeepMind's AlphaGenome Atlas predicts impact of all 9 billion possible DNA letter changes — Google DeepMind · 2026-09-08
- Study of 26,811 Chinese Students: AI Boosts Homework Scores 18% but Cuts Exam Results 20% — soumitrashukla9 · 2026-09-08
- LeVJEPA matches V-JEPA 2 with up to 20x less pretraining compute, no heuristics — Cohere_Labs · 2026-09-08
- McGill txtlab releases 149K plot summaries and crowd-sourced narrative datasets — _akpiper · 2026-09-08
- Most recurring multilingual SAE features fail causal validation — only one truly drives translation in Gemma — ravithejads · 2026-09-08
- Spectrogram-to-Sound AI Meets Bioacoustics for Deeper Bird Vocalization Research — CatAstro_Piyush · 2026-09-08