Loading...
Market
What will be the performance improvement of GoogleDeepMind's IRL method over traditional methods by end of 2024?
2
Resolution / Starting Odds
Less than 10% • 25%
10% to 20% • 25%
20% to 30% • 25%
More than 30% • 25%
Benchmark results published by GoogleDeepMind or a reputable AI research organization
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
Yes • 50%
No • 50%
MATH dataset • 25%
Natural Language Processing • 25%
Computer Vision • 25%
Other • 25%
GLUE • 25%
SuperGLUE • 25%
SQuAD • 25%
Other • 25%
Less than 5% • 25%
5% to 10% • 25%
10% to 20% • 25%
More than 20% • 25%
Improved win rate against intermediate players • 25%
Enhanced adaptability to different playing styles • 25%
Reduced reaction time • 25%
Other • 25%
25% gain in MATH • 25%
15% gain in other domains • 25%
Integration into multiple commercial products • 25%
Other • 25%
New AI system for medical research • 25%
New AI system for financial modeling • 25%
New AI system for climate modeling • 25%
Other • 25%
25% gain in self-correction for MATH dataset • 25%
15% gain in self-correction for other datasets • 25%
Adoption by three major tech companies • 25%
Other • 25%
Yes • 50%
No • 50%
Beginner • 25%
Intermediate • 25%
Advanced • 25%
Professional • 25%
Yes • 50%
No • 50%
