Palette trains a small model to build branded decks

Focus: Applied Generative AI for Knowledge Workers
Date: November 3, 2026
Topics: Presentation design, Small models, Visual QA
Evidence: Publisher Authored
Availability: Available

Structured brief feeds slide code, render and geometry checks.

Summary: IBM Research describes Palette, a 20-billion-parameter open model trained on about 1,000 presentations and 10,000 slides. A designer component emits structured plans, a coder writes PptxGenJS, and automated overlap and overflow checks trigger repairs. On PresentBench, IBM reports the checking harness improved results by 3.5 points.

Why it matters: Presentation automation becomes more useful when layout errors are treated as test failures instead of cosmetic surprises. Teams can borrow the planner–coder–checker pattern, but should evaluate brand fidelity, accessibility and complex-chart handling on their own deck library.

Source: Palette: A Small Open Model for Creating Presentation Slides

How useful was this?

Anonymous feedback. No name or email collected.

← Daily Brief for November 4, 2026