Добавить в корзинуПозвонить
Найти в Дзене
Crynet.io

🖥 NVIDIA just dropped their 7B size Cosmos Reason reasoning-VLM for Physical AI

🖥 NVIDIA just dropped their 7B size Cosmos Reason reasoning-VLM for Physical AI! What’s the deal? • It’s an open and customizable reasoning-VLM for physical AI & robotics: the model understands video + text, “thinks” step-by-step, and makes real-world decisions. 😎 What can it do? 👀 See, reason, and act in the real world 🛠 Tackle multi-step tasks and handle tricky/new situations 🤖 Perfect for developing robots, autonomous vehicles, and visual AI agents. Why is it a big deal? • Post-tuning gives a serious boost: SFT adds about +10% to the base, RL brings in another ≈ +5%. • The result? An average score of about 65.7 on key benchmarks for robots and autopilot! 🚀📊

🖥 NVIDIA just dropped their 7B size Cosmos Reason reasoning-VLM for Physical AI!

What’s the deal?

• It’s an open and customizable reasoning-VLM for physical AI & robotics: the model understands video + text, “thinks” step-by-step, and makes real-world decisions. 😎

What can it do?

👀 See, reason, and act in the real world

🛠 Tackle multi-step tasks and handle tricky/new situations

🤖 Perfect for developing robots, autonomous vehicles, and visual AI agents.

Why is it a big deal?

• Post-tuning gives a serious boost: SFT adds about +10% to the base, RL brings in another ≈ +5%.

• The result? An average score of about 65.7 on key benchmarks for robots and autopilot! 🚀📊