Loading...
Market
In which application will GoogleDeepMind's IRL method be used by end of 2024?
2
Resolution / Starting Odds
Chatbots • 25%
Translation Services • 25%
Content Generation • 25%
Other • 25%
Official announcements or press releases from GoogleDeepMind
Story
GoogleDeepMind Introduces Scalable Inverse Reinforcement Learning for Language Models
Sep 5, 2024, 09:23 AM
DeepMind has introduced a new approach to language model training using Scalable Inverse Reinforcement Learning (IRL). This method presents an effective alternative to traditional supervised Maximum Likelihood Estimation (MLE) in the fine-tuning pipeline, resulting in more robust reward functions and increased performance and diversity of model generations. The foundation of this approach lies in imitation learning, which is considered a reinforcement learning problem. Compared to supervised learning, IRL better exploits sequential structure, online data, and further extracts rewards. The insights were shared in a recent paper by GoogleDeepMind.
View original story
Similar markets
Mathematics • 25%
Healthcare • 25%
Natural Language Processing • 25%
Other • 25%
Healthcare AI • 25%
Natural Language Processing • 25%
Autonomous Systems • 25%
Other • 25%
Education • 25%
Healthcare • 25%
Finance • 25%
Other • 25%
Computer Vision • 25%
Natural Language Processing • 25%
Robotics • 25%
Other • 25%
Coding/Programming • 25%
Academic Research • 25%
Business Analytics • 25%
Other • 25%
Deep Learning • 25%
Computer Vision • 25%
Autonomous Vehicles • 25%
Robotics • 25%
Healthcare AI systems • 25%
Autonomous vehicles • 25%
Financial trading algorithms • 25%
Customer service chatbots • 25%
Google Search • 25%
Google Assistant • 25%
Google Cloud AI • 25%
Other • 25%
Healthcare • 25%
Finance • 25%
Education • 25%
Other • 25%
Customer Service • 25%
Data Analysis • 25%
Research Assistance • 25%
Other • 25%
Healthcare • 25%
Finance • 25%
Education • 25%
Other • 25%