Google shows how to turn DiffusionGemma into a scorer via vLLM templates
bodonoghue85 · x · 2026-09-30
Google's Gemma team shared a trick to turn DiffusionGemma into a Jev-like model: seed the canvas with a response template in vLLM, and you can extract confidence scores and probability distributions for yes/no, multiple-choice, and scored questions in a single denoising step — fast, efficient, and ready to scale for structured tasks.
More from coding & agent
- Using Copilot CLI to set up a Mac-to-Windows devtunnel for cross-machine testing — DanWahlin · 2026-09-30
- Microsoft Research finds LLMs show Dunning-Kruger-style overconfidence in coding — burkov · 2026-09-30
- Integrating a full app into ChatGPT via Plugin Extensions to run social posting — haltakov · 2026-09-30
- Seroter's reading list: AI-picked dependencies, faster AI coding means harder engineering — rseroter · 2026-09-30
- Microsoft launches Azure canvases: manage Azure resources inside GitHub Copilot — DanWahlin · 2026-09-30
- Much of Anthropic's internal software is now built by Claude, says Mike Krieger — victor_explore · 2026-09-30