daily-hour-news·
🔬IBM's Palette: 20B Model Builds Decks, 1.1 Behind NotebookLM
TL;DR
IBM Research fine-tuned a 20B model with a LoRA adapter on about 1,000 synthetic decks to build slides in a house design language. On PresentBench it ranks second on the three-domain mean, 59.7 against NotebookLM's 60.8.
IBM Research fine-tuned a 20B model with a LoRA adapter on about 1,000 synthetic decks to build slides in a house design language. On PresentBench it ranks second on the three-domain mean, 59.7 against NotebookLM's 60.8.
Key Points
1
Training data: about 1,000 synthetic decks, roughly 10,000 slides
Keep reading — it's free
Enter your email to keep reading — plus the best of AI & tech, daily. Free, forever.
Also get
or
Already a member? Sign in
Quick Facts
IBM ResearchPaletteLoRAslide generationPresentBenchNotebookLMsynthetic data
Comments
Subscribe to join the conversation...
Be the first to comment