The story’s score and claims

Comments on No. 73 of 142 · shuklabhay.github.io · edition of Wednesday 18 November 2026

How image generation got good (diffusion derivations)

Summary

This post explains why image generation became reliable, focusing on diffusion models. It argues that L2 regression blurs outputs because it averages over multimodal possibilities, while diffusion iteratively denoises toward a specific sample.

Comments

Ranked by a public rubric, not by votes: what a comment adds counts, not who wrote it. How comments are ranked · Comment rules