MIRO: MultI-Reward cOnditioned pretraining improves T2I quality and efficiency

A darkroom with one orange safelight: every badge, number and rule glows the same warm tangerine against near-black.

nicolas-dufour.github.io/miroMIRO: MultI-Reward cOnditioned pretraining improves T2I quality and efficiency
Screenshot of https://nicolas-dufour.github.io/miro, © the authors. Shown for reference. Request removal
Venue
ICML 2026
Year
2026
First author
Nicolas Dufour
Field
text-to-image generation
Template
custom (styles.css with KaTeX, d3, p5.js)
Theme
Dark

Checked against the rendered page. KPI numbers count up on scroll; the screenshot shows intermediate values (13x, 247x) while the markup's final values are 19x and 370x. No web font is loaded; the system UI face renders. A light theme exists behind a toggle.

Why it is here

Dark theme with a consistent orange accent through the badge, stats and heading bars. Tiered stat cards, an image carousel and a chip-style contents list give it a confident blog-like presentation.

Overview

Color Palette

Typography

system-ui
MIRO: MultI-Reward cOnditioned pretraining improves T2I quality and ef Abstract · Method · Results · BibTeX — 0123456789
Courier New
MIRO: MultI-Reward cOnditioned pretraining improves T2I quality and ef Abstract · Method · Results · BibTeX — 0123456789
KaTeX
MIRO: MultI-Reward cOnditioned pretraining improves T2I quality and ef Abstract · Method · Results · BibTeX — 0123456789

Spacing & Shape

Page Modules

Guidelines

More like this