微软研究院博客

Dion: the distributed orthonormal update revolution is here

2025年8月12日 | Kwangjun AhnJohn Langford
Dion is a new AI model optimization method that boosts scalability and performance over existing leading methods by orthonormalizing only a top rank subset of singular vectors, enabling more efficient training of large models such as LLaMA-3 with reduced overhead.

最新文章

  1. Stylized digital illustration of a multi-layered circuit board. A glowing blue microchip sits at the top center, with intricate circuitry radiating outward. Beneath it, four stacked layers transition in color from blue to orange, each featuring circuit-like patterns. Smaller rectangular and circular components are connected around the layers, all set against a dark background with scattered geometric shapes.

    Project Ire autonomously identifies malware at scale 

    2025年8月5日

    Designed to classify software without context, Project Ire replicates the gold standard in malware analysis through reverse engineering. It streamlines a complex, expert-driven process, making large-scale malware detection faster & more consistent.

  2. CollabLLM blog hero | flowchart diagram starting in the upper left corner with an icon of two overlapping chat bubbles; arrow pointing right to an LLM network node icon; branching down to show three simulated users; right arrow to a "Reward" box

    CollabLLM: Teaching LLMs to collaborate with users 

    2025年7月15日

    Recipient of an ICML 2025 Outstanding Paper Award, CollabLLM improves how LLMs collaborate with users, including knowing when to ask questions and how to adapt tone and communication style to different situations. This approach helps move AI toward more user-centric and trustworthy systems.

Explore More