SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2303.13703 · ICCV · 2023

End-to-End Diffusion Latent Optimization Improves Classifier Guidance

Stefano Ermon, Nikhil Naik, Wallace Bram, A Salesforce

arXiv · PDF · Open in the Atlas

Code that ran

We have not lifted any functions out of this paper's repositories yet, so there is nothing we have run. The repositories linked to it are listed below.

Repositories linked to this paper

Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.

Abstract

A dog swimming" "Darth Vader playing with a raccoon on Mars during sunset" CLIP Text Encoder Aesthetic Scoring Model CLIP Visual Encoder "A Blue-Winged Warbler" Figure 1: We propose DOODL -a process that directly optimizes diffusion latents w.r.t. a model-based loss on the final generation. Our method improves on vanilla classifier guidance in all tested settings and we demonstrate capabilities novel to this class of methods such as vocabulary expansion, entity personalization, and perceived aesthetic value improvement.

For agents

The same record, over MCP at https://syntology.ai/mcp:

get_harvested_code_for_paper("2303.13703")
get_code_for_paper("2303.13703")
have("2303.13703")

The run record, dated, one paper per request, free:

curl https://syntology.ai/api/ran/2303.13703.json

A badge for a README (the split and the date, never a ratio):

[![Syntology run record](https://syntology.ai/api/ran/2303.13703.svg)](https://syntology.ai/paper/2303.13703)

Connect an agent — have() is free.