Telegram Group & Telegram Channel
This media is not supported in your browser
VIEW IN TELEGRAM
Еще интересный подход по созданию агентов (на базе Vision language модели) с RLем которые могут пользоваться Android телефоном через GUI

Вначале трейнят offline RL на данных, потом offline-to-online где агент уже сам доучивается в среде. Создали распараллеленный симулятор который одновременно может запускать 64 эмулятора андроида.

Тестировались на датасете Android-in-the-Wild (AitW). VLMка на 1.3B параметров.

* success rate подняли до 67.2%

у другого RL агента который учился через Behavior cloning был - 57.8%

GPT-4V - 8.3%
Gemini 1.5 Pro - 17.7%
17B CogAgent - 38.5%

DigiRL: Training In-The-Wild Device-Control Agents with Autonomous Reinforcement Learning
https://arxiv.org/abs/2406.11896

https://digirl-agent.github.io/
https://github.com/DigiRL-agent/digirl
🔥7👍2



group-telegram.com/AGI_and_RL/800
Create:
Last Update:

Еще интересный подход по созданию агентов (на базе Vision language модели) с RLем которые могут пользоваться Android телефоном через GUI

Вначале трейнят offline RL на данных, потом offline-to-online где агент уже сам доучивается в среде. Создали распараллеленный симулятор который одновременно может запускать 64 эмулятора андроида.

Тестировались на датасете Android-in-the-Wild (AitW). VLMка на 1.3B параметров.

* success rate подняли до 67.2%

у другого RL агента который учился через Behavior cloning был - 57.8%

GPT-4V - 8.3%
Gemini 1.5 Pro - 17.7%
17B CogAgent - 38.5%

DigiRL: Training In-The-Wild Device-Control Agents with Autonomous Reinforcement Learning
https://arxiv.org/abs/2406.11896

https://digirl-agent.github.io/
https://github.com/DigiRL-agent/digirl

BY Агенты ИИ | AGI_and_RL


Share with your friend now:
group-telegram.com/AGI_and_RL/800

View MORE
Open in Telegram


Telegram | DID YOU KNOW?

Date: |

"There are several million Russians who can lift their head up from propaganda and try to look for other sources, and I'd say that most look for it on Telegram," he said. Lastly, the web previews of t.me links have been given a new look, adding chat backgrounds and design elements from the fully-features Telegram Web client. Andrey, a Russian entrepreneur living in Brazil who, fearing retaliation, asked that NPR not use his last name, said Telegram has become one of the few places Russians can access independent news about the war. In a statement, the regulator said the search and seizure operation was carried out against seven individuals and one corporate entity at multiple locations in Ahmedabad and Bhavnagar in Gujarat, Neemuch in Madhya Pradesh, Delhi, and Mumbai. These administrators had built substantial positions in these scrips prior to the circulation of recommendations and offloaded their positions subsequent to rise in price of these scrips, making significant profits at the expense of unsuspecting investors, Sebi noted.
from tw


Telegram Агенты ИИ | AGI_and_RL
FROM American