<?xml version="1.0" encoding="UTF-8"?><oembed><type>video</type><version>1.0</version><html>&lt;iframe src=&quot;https://www.loom.com/embed/44c793c47e7d45eaaf02bac7c168a10d&quot; frameborder=&quot;0&quot; width=&quot;1440&quot; height=&quot;1080&quot; webkitallowfullscreen mozallowfullscreen allowfullscreen&gt;&lt;/iframe&gt;</html><height>1080</height><width>1440</width><provider_name>Loom</provider_name><provider_url>https://www.loom.com</provider_url><thumbnail_height>1080</thumbnail_height><thumbnail_width>1440</thumbnail_width><thumbnail_url>https://cdn.loom.com/sessions/thumbnails/44c793c47e7d45eaaf02bac7c168a10d-4c58c90bdfbb77cb.gif</thumbnail_url><duration>77.612</duration><title>NOUS Hackaton - Presentation - 18 May 2025</title><description>In this video, I discuss the importance of fine-tuning larger gauge models to interact effectively with MCPs, which are crucial for intelligent internets. I propose a reinforcement learning framework to help LLMs master MCPs by generating datasets and evaluating use cases. I also highlight the challenges developers face with MSW API calls and the diversity of MCPs. Please take a moment to consider the proposed framework and how it could enhance our projects.</description></oembed>